[opensuse-packaging] why there are so few packages in Factory, revisited
Hello, you might remember recent discussion in opensuse-factory mailing list on this topic. While the statistics that started it were inherently flawed, there seemed to be a consensus that Factory doesn't have as many packages as it could and that there are many packages in devel projects or home projects that are never submitted to Factory. I considered sending my opinion why but I was afraid it might start a flame without any positive outcome. Recent experience conviced me to try anyway. Disclaimer: the sole purpose of this e-mail is to present my view on the problem and to give responsible people something to think about. It is definitely not meant as a complaint about the particular case (which serves only as an example) or some kind of pointing fingers. Short version: I believe main reason why too many packages never move from home or devel projects to Factory is that on one hand, OBS made it extremely easy to build a package for a set of distributions and even add your own repository with such packages; on the other hand, once you decide to get the Factory, you have to tackle constantly increasing amount of "bureaucracy" to even get the package in and then face a stream of pointless changes in the name of some high standards which are apparently constantly changing and actually not even universally accepted at any given moment. There is a popular phrase about "scratching an itch". The problem, as I see it, is that to scratch one's itch, home project is sufficient. If you are an altruist, you might even push the package into a devel project so that it stands out in the search. But getting it into Factory, that's no more about scratching your itch; you must have really strong motivation for that - and even stronger to do it again once you know what you are going to face. Let's look at an example. Few years ago, I discovered tcpreplay, a set of network utilities which are quite handy when investigating networking issues. I created a package, did my best to avoid OBS warnings and put it into a project under my home so that I could easily install it on my machines. That scratched my itch. Later some guy created a submit request for network:utilities; I checked the package, updated it to new versions, did some cleanup and package ended up in network:utilities. After the December discussion, I decided to submit the tcpreplay to Factory. I checked the build, polished the specfile a bit and submitted it. The request was declined by a check script because a copyright was missing. OK, fair enough, while I don't consider that particular specfile such a great piece of art to need a copyright, why not. I ran the specfile through the spec cleaner service (which often happens automatically with "osc build" but for some reason this wasn't the case). It did some rearrangement and style changes; some I didn't quite like but nothing crucial, so I resubmitted the package. With that submit request still waiting in staging repo, on Sunday I got a request against network:utilities. I didn't have time to review it so I left it to the morning. Before I could do so, 5 hours after creation of the request, to my surprise, it was accepted by the guy who originally poked me to submit the package to network:utilities (but never contributed to the package in any way since). I checked the changes anyway (after all, I'm bugowner so I feel responsible for the package) and I really don't know what to think: https://build.opensuse.org/package/rdiff/network:utilities/tcpreplay?linkrev=base&rev=5 - replace xz compressed tarball by the original gzipped one For years I've been told by OBS that I should repack gzipped tarballs to save resources. It's not true any more? OK, it allows for checking the signature so it makes sense. - make BuildRoot specification unconditional For years I've been told that it is obsolete. The only reason I keep it is to allow build for SLE11. Why forcing it on distributions that don't need it? - replace source files by full URL I don't like this change. More precisely: I hate it. - add "-q" to %setup Why? No idea. - add "V=1" to make Why? To make build warnings harder to see? - break configure line into two If it needed breaking, I would do so. This one had 32 characters... Seriously? - replace ${RPM_BUILD_ROOT} with %{buildroot} The only rule I have ever seen about this was there is no rule. - add %{?_smp_mflags} also to make install Well, why not. But definitely not _after_ the argument. - add LICENSE to the package OK, why not. - %files: expand wildcards to explicit lists Some people apparently prefer package build to fail whenever a file is added or renamed. I don't force them not to do so in their packages, could they respect my preferences in mine? Apparently not. - reorder tags in specfile header I'm confused. The previous order was what spec cleaner did. Wasn't it supposed to be The Right Order(TM)? And what happens next time spec cleaner service is run? Is it going to change the order back, masking real changes? Summary: lot of meaningless churn, one useful change, two I dislike and two (or two and a half) I really dislike. And a bad feeling that while I took responsibility for the package, I apparently have no say in what should its specfile look like. Don't understand me wrong: I have no problem with reasonable rules, even if they are quite strict (like the tagging policy in our kernel packages or strict coding style policy in kernel networking). But here I feel this crosses a line of harassment. OK, I'll learn to live with it. I'll bite the bullet with packages like firebird or twinkle which are kind of "heart thing". But next time I stumble upon a package that might be useful, I'll think twice before submitting it to Factory. Had I known what was going to happen, well, there would be one less package in Factory (assuming tcpreplay is going to make it). And just today, I got a reaction like "Been there. I know exactly what you are talking about." Sure, you may say that it's my fault I don't understand openSUSE high standards. You might even say openSUSE is not interested in packagers who don't understand its high standards. But you should understand in the end it's always packager's choice if he is going to submit the package to Factory or not. And remember, to "scratch an itch", putting it into a home project is enough in most cases; submitting to Factory is not about "scratching an itch", it's about helping others; things like this work as an active discouragement. Think about it next time you ask yourself why so many packages stay in home or devel project without ever being submitted to Factory. Or, even better, think about it next time you feel urge to completely rearrange a package you are not maintaining (and not going to) just to force your aesthetic criteria on it (and its maintainer). Michal Kubeček -- To unsubscribe, e-mail: opensuse-packaging+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-packaging+owner@opensuse.org
Michal, et.al. Let me first tell you: I understand your frustration. And the fact that somebody with a project maintainer role oversteps the package maintainer, giving him 5 hours of notice period, is bad behavior. I stronlgy encourage the two of you to talk this out directly. As to the points you list in the 'change set', I only want to give some small words there (from a opensuse-review team perspective - but it's my own, which I hope the rest of the reviewer team shares). On Tue, 2016-02-02 at 00:10 +0100, Michal Kubecek wrote:
- replace xz compressed tarball by the original gzipped one For years I've been told by OBS that I should repack gzipped tarballs to save resources. It's not true any more? OK, it allows for checking the signature so it makes sense.
Most of the distro packages switched to using the prisitine upstream tarball quite a while ago, especially for checking signatures, and, as I will explain in pioint 3
- make BuildRoot specification unconditional For years I've been told that it is obsolete. The only reason I keep it is to allow build for SLE11. Why forcing it on distributions that don't need it?
It is indeed needed for SLE11; I have no strong pro or contra if it should be provided for non-SLE11 builds, but it's purely used during the build phase, so does not impact the resulting package. So it comes down to: do you prefer a condition without added value in your spec file? Or do you prefer a 'plain' readable one with as few churn as possible. As you stated: it's your package - your choice.
From a Factory Reviewer PoV: no objection either way. YOUR CHOICE!
- replace source files by full URL I don't like this change. More precisely: I hate it.
It has one big advantage: trust! Every package submitted to openSUSE:Factory, the review bot actually downloads the tarball from the URL in the Source line and verifies that the submitted one actually IS The one it's supposed to be. This avoids that a packager can inject anything into a tarball (as you can imagine, there are not sufficient resources to manually review all the submitted code - we need to be able to trust one another. The URL as Source allows to automate it)/
- add "-q" to %setup Why? No idea.
Again, personal choice. Factory Reviewer accepts either way. The most notable thing '-q' does here is not list the entire content of the extracted tarball... so making the build log 'neater'
- add "V=1" to make Why? To make build warnings harder to see?
A constant battle - I know that one. I myself consider V=1 great for debug perpose... as long as gcc's warnings are there if it's not specified, I don't enable it.
- break configure line into two If it needed breaking, I would do so. This one had 32 characters... Seriously?
That's really a strange one to even think to enforce. Certainly, no packaging guideline exists around this topic. And obviously, there are cases it makes sense to split the lines, there are others it's useless. Your case seems to be the latter.
- replace ${RPM_BUILD_ROOT} with %{buildroot} The only rule I have ever seen about this was there is no rule.
There is a small difference in the expansion of the variable: does bash do it (${}) or does rpm do it %{}. The value is the same. style-wise, many prefer the lower case writing style (spec-cleaner does change this one to).
- add %{?_smp_mflags} also to make install Well, why not. But definitely not _after_ the argument.
Parallel installation - how big is that package :) I doubt it makes much difference (but then: the latest version of spec-cleaner does this automatically IIRC - so the submitter might not have done this explicitly)
- add LICENSE to the package OK, why not.
Actually mandatory to pass legal review. Which, by the way, is what the current submission to openSUSE:Factory is pending at: https://build.ope nsuse.org/request/show/356511 (not staging per se.... just that it can only move out of staging once all review steps are handled)
- %files: expand wildcards to explicit lists Some people apparently prefer package build to fail whenever a file is added or renamed. I don't force them not to do so in their packages, could they respect my preferences in mine? Apparently not.
Indeed - personal style to chose. From experience, I use explicit lists when it's about a lot of plugin-style files that get auto-enabled based on deps. Like this, I see when a dep goes missing: the build fails.
- reorder tags in specfile header I'm confused. The previous order was what spec cleaner did. Wasn't it supposed to be The Right Order(TM)? And what happens next time spec cleaner service is run? Is it going to change the order back, masking real changes?
Sometimes, new version of spec cleaner might do it different. Only solution is for us all to run the same version of the tool.
Summary: lot of meaningless churn, one useful change, two I dislike and two (or two and a half) I really dislike. And a bad feeling that while I took responsibility for the package, I apparently have no say in what should its specfile look like. Don't understand me wrong: I have no problem with reasonable rules, even if they are quite strict (like the tagging policy in our kernel packages or strict coding style policy in kernel networking). But here I feel this crosses a line of harassment.
I agree that there are quite some questionable changes forced into your package. As an active maintainer, I actually suggest you to REVERT the change and implement the changes you like on your own. Don't play around: You are the maintainer of the package, you care for it. Show it - revert.
OK, I'll learn to live with it. I'll bite the bullet with packages like
Please don't just learn to live with it! Your approach to bring this up here is absolutely perfect and I hope peopple take this constructive critic on when they do changes. Do not step on each others feet, but collaborate!
firebird or twinkle which are kind of "heart thing". But next time I stumble upon a package that might be useful, I'll think twice before submitting it to Factory. Had I known what was going to happen, well, there would be one less package in Factory (assuming tcpreplay is going to make it). And just today, I got a reaction like "Been there. I know exactly what you are talking about."
In the hopes not to lose your spirit of helping others with your work: please re-consider... your packages are welcome in the distribution.
Sure, you may say that it's my fault I don't understand openSUSE high standards. You might even say openSUSE is not interested in packagers who don't understand its high standards. But you should understand in the end it's always packager's choice if he is going to submit the package to Factory or not. And remember, to "scratch an itch", putting it into a home project is enough in most cases; submitting to Factory is not about "scratching an itch", it's about helping others; things like this work as an active discouragement.
Whereas it's true that we have packaging guidelines, and some rules we enforce stricter than others: ever tried to submit something to Debian? :) Jokes aside: our guidelines are there to 'guide' - some of the rules are strictly enforced (most of them by a bot!), the review team tries to 'see more common errors' and things the bot might mis-understand (like: static libs: they are a no-go, except when done properly and documented and fully considered - so the bot can't know)
Think about it next time you ask yourself why so many packages stay in home or devel project without ever being submitted to Factory. Or, even better, think about it next time you feel urge to completely rearrange a package you are not maintaining (and not going to) just to force your aesthetic criteria on it (and its maintainer).
Thank you Michal for bringing this up! I agree: aestethics are not a reason for a decline (even though I can assure you that a spec that looks ugly and is more cryptic to read than needed will likely make it to the 'end of the review queue'). For me, the important message in this entire thing is to ALL contributors: contribute to any package you want - be active, but if there is a person dedicated to maintain it (and this person is still active): let this person have a word if he/she likes the change. There are 'teams' building around certain packages, where people talk together and come to a common view on how they want to maintain it - so they jointly do the work, sometimes with, sometimes without consulting. There are even larger projects where NO maintainer in the project is supposed to accept his own change! Strict 4-eye peer review even before stuff moves to Factory. This is not mandatory in OBS - but if the active maintainer so wishes, it is to be respected. Cheers, Dominqiue -- To unsubscribe, e-mail: opensuse-packaging+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-packaging+owner@opensuse.org
On Tue, Feb 02, 2016 at 12:49:46AM +0100, Dominique Leuenberger / DimStar wrote:
Michal, et.al.
- add "V=1" to make Why? To make build warnings harder to see?
A constant battle - I know that one. I myself consider V=1 great for debug perpose... as long as gcc's warnings are there if it's not specified, I don't enable it.
Actually we are parsing the gcc commandline to spot if you are using optflags and do some post build checks. (post-build-checks, check_gcc)
- add %{?_smp_mflags} also to make install Well, why not. But definitely not _after_ the argument.
Parallel installation - how big is that package :) I doubt it makes much difference (but then: the latest version of spec-cleaner does this automatically IIRC - so the submitter might not have done this explicitly)
I fixed several packages that build fine with -j2 but did not with -j8, which was annoying. so this actually has the cause to break things, but it is rare. Ciao, Marcus -- To unsubscribe, e-mail: opensuse-packaging+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-packaging+owner@opensuse.org
On úterý 2. února 2016 8:35:56 CET Marcus Meissner wrote:
On Tue, Feb 02, 2016 at 12:49:46AM +0100, Dominique Leuenberger / DimStar wrote:
Michal, et.al.
- add "V=1" to make Why? To make build warnings harder to see?
A constant battle - I know that one. I myself consider V=1 great for debug perpose... as long as gcc's warnings are there if it's not specified, I don't enable it.
Actually we are parsing the gcc commandline to spot if you are using optflags and do some post build checks. (post-build-checks, check_gcc) Indeed there are actually several cases when lack of of usage of optflags was not observed due to silent build, such as in ninja, thin-provisioning tools and himeno - and those are just few cases which I noticed in last few days.
- add %{?_smp_mflags} also to make install Well, why not. But definitely not _after_ the argument.
Parallel installation - how big is that package :) I doubt it makes much difference (but then: the latest version of spec-cleaner does this automatically IIRC - so the submitter might not have done this explicitly)
I fixed several packages that build fine with -j2 but did not with -j8, which was annoying. so this actually has the cause to break things, but it is rare.
Ciao, Marcus
On Tuesday 02 of February 2016 00:49:46 Dominique Leuenberger / DimStar wrote:
Let me first tell you: I understand your frustration. And the fact that somebody with a project maintainer role oversteps the package maintainer, giving him 5 hours of notice period, is bad behavior. I stronlgy encourage the two of you to talk this out directly.
To be precise, the request was accepted by someone else. But in general, people not actively maintaining a package accepting requests without giving maintainers enough time to react, is kind of pattern I see too often to my liking.
As to the points you list in the 'change set', I only want to give some small words there (from a opensuse-review team perspective - but it's my own, which I hope the rest of the reviewer team shares).
On Tue, 2016-02-02 at 00:10 +0100, Michal Kubecek wrote:
- replace xz compressed tarball by the original gzipped one For years I've been told by OBS that I should repack gzipped tarballs to save resources. It's not true any more? OK, it allows for checking the signature so it makes sense.
Most of the distro packages switched to using the prisitine upstream tarball quite a while ago, especially for checking signatures, and, as I will explain in pioint 3
I guess it wasn't completely clear from my previous mail: this particular change I understand and appreciate. I just wanted to list even the changes I have no objections to so that it didn't look like I disliked everything. Actually, I never liked the policy of recompressing (or even repacking) upstream tarballs.
- make BuildRoot specification unconditional For years I've been told that it is obsolete. The only reason I keep it is to allow build for SLE11. Why forcing it on distributions that don't need it?
It is indeed needed for SLE11; I have no strong pro or contra if it should be provided for non-SLE11 builds, but it's purely used during the build phase, so does not impact the resulting package. So it comes down to: do you prefer a condition without added value in your spec file? Or do you prefer a 'plain' readable one with as few churn as possible. As you stated: it's your package - your choice.
As it does no harm (and it's unlikely to in a foreseeable future), I take the conditional mostly as a reminder that it's temporary and it's going away one day. I guess a comment would be sufficient for that purpose.
- replace source files by full URL I don't like this change. More precisely: I hate it.
It has one big advantage: trust! Every package submitted to openSUSE:Factory, the review bot actually downloads the tarball from the URL in the Source line and verifies that the submitted one actually IS The one it's supposed to be. This avoids that a packager can inject anything into a tarball (as you can imagine, there are not sufficient resources to manually review all the submitted code - we need to be able to trust one another. The URL as Source allows to automate it)/
OK, this changes things. If the URL is actually used for something meaningful, I have no problem with using it.
- add LICENSE to the package OK, why not.
Actually mandatory to pass legal review.
I was afraid so. While I find the overall obsession with licenses a bit sad, I understand that (open)SUSE has to live with it and watch its back.
- reorder tags in specfile header I'm confused. The previous order was what spec cleaner did. Wasn't it supposed to be The Right Order(TM)? And what happens next time spec cleaner service is run? Is it going to change the order back, masking real changes?
Sometimes, new version of spec cleaner might do it different. Only solution is for us all to run the same version of the tool.
Perhaps changing the preferred order only if there is a really strong reason to might help a bit.
I agree that there are quite some questionable changes forced into your package. As an active maintainer, I actually suggest you to REVERT the change and implement the changes you like on your own. Don't play around: You are the maintainer of the package, you care for it. Show it - revert.
I'll revert the changes I object to and probably keep those I only find useless but the result does not particularly bother me (e.g. the tags order).
For me, the important message in this entire thing is to ALL contributors: contribute to any package you want - be active, but if there is a person dedicated to maintain it (and this person is still active): let this person have a word if he/she likes the change. There are 'teams' building around certain packages, where people talk together and come to a common view on how they want to maintain it - so they jointly do the work, sometimes with, sometimes without consulting. There are even larger projects where NO maintainer in the project is supposed to accept his own change! Strict 4-eye peer review even before stuff moves to Factory. This is not mandatory in OBS - but if the active maintainer so wishes, it is to be respected.
Agreed. And I'm glad to hear that. Michal Kubeček -- To unsubscribe, e-mail: opensuse-packaging+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-packaging+owner@opensuse.org
* Dominique Leuenberger / DimStar (dimstar@opensuse.org) [20160202 00:51]:
Indeed - personal style to chose. From experience, I use explicit lists when it's about a lot of plugin-style files that get auto-enabled based on deps. Like this, I see when a dep goes missing: the build fails.
I used to prefer wildcards but then the build of one binary failed unnoticed and we released the package it's part of. Subsequently a bug report was opened when a user noticed the omission. After that I changed to explicitely list the files in my packages. Philipp -- To unsubscribe, e-mail: opensuse-packaging+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-packaging+owner@opensuse.org
On Fri, 2016-02-05 at 13:29 +0100, Philipp Thomas wrote:
* Dominique Leuenberger / DimStar (dimstar@opensuse.org) [20160202 00:51]:
Indeed - personal style to chose. From experience, I use explicit lists when it's about a lot of plugin-style files that get auto-enabled based on deps. Like this, I see when a dep goes missing: the build fails.
I used to prefer wildcards but then the build of one binary failed unnoticed and we released the package it's part of. Subsequently a bug report was opened when a user noticed the omission. After that I changed to explicitely list the files in my packages.
Thanks for sharing this Philipp, It confirms that what is often perceived as 'less work' can turn out as more work/frustration later on when you have to chase a bug that could have been avoided. It's a trade-off the packager has to chose - and live with the consequences. Dominique -- To unsubscribe, e-mail: opensuse-packaging+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-packaging+owner@opensuse.org
Am 05.02.2016 um 13:36 schrieb Dominique Leuenberger / DimStar:
On Fri, 2016-02-05 at 13:29 +0100, Philipp Thomas wrote:
* Dominique Leuenberger / DimStar (dimstar@opensuse.org) [20160202 00:51]:
Indeed - personal style to chose. From experience, I use explicit lists when it's about a lot of plugin-style files that get auto-enabled based on deps. Like this, I see when a dep goes missing: the build fails.
I used to prefer wildcards but then the build of one binary failed unnoticed and we released the package it's part of. Subsequently a bug report was opened when a user noticed the omission. After that I changed to explicitely list the files in my packages.
Thanks for sharing this Philipp,
It confirms that what is often perceived as 'less work' can turn out as more work/frustration later on when you have to chase a bug that could have been avoided.
I had a similar issue in the qemu package where binaries and firmware files ended up in an unintended sub-package. We switched to an explicit file list, so that we notice new files on version update. It didn't really break anything but it made a common package much larger, packages inconsistent and now requires carrying a "Provides: originalpackage:filepath" as remedy. The downside is that the package build takes quite some time until those errors show up.
It's a trade-off the packager has to chose - and live with the consequences.
Cheers, Andreas -- SUSE Linux GmbH, Maxfeldstr. 5, 90409 Nürnberg, Germany GF: Felix Imendörffer, Jane Smithard, Graham Norton; HRB 21284 (AG Nürnberg) -- To unsubscribe, e-mail: opensuse-packaging+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-packaging+owner@opensuse.org
On Tuesday 2016-02-02 00:10, Michal Kubecek wrote:
Short version: I believe main reason why too many packages never move [...] to Factory is that [...] you have to tackle constantly increasing amount of "bureaucracy" to even get the package in
If you think openSUSE suffers from a bureaucracy problem, just wait until you look elsewhere. Fedora: https://fedoraproject.org/wiki/Join_the_package_collection_maintainers Let's just work down that list, and compare with openSUSE: * no hard requirement to join mailing lists * no hard requirement to inform upstream * no need for the sponsoring game (but maybe that's implicit in the review in openSUSE) * reviews usually go a lot faster than what I can find on [ https://fedoraproject.org/PackageReviewStatus/NEEDSPONSOR.html ], feels like a lot less bikeshedding in openSUSE:Factory. Debian: https://www.debian.org/devel/join/ https://www.debian.org/devel/join/newmaint * No "application", no "recommendation", no "membership" needed In a way, openSUSE is more like github. Register a (technical) account, create/fork, submit, and pretty much done.
Let's look at an example. [...] tcpreplay
- replace xz compressed tarball by the original gzipped one For years I've been told by OBS [...]
That's now years away. 4 maybe? Please don't tell me you just joined SUSE.
- make BuildRoot specification unconditional For years I've been told that it is obsolete. The only reason I keep it is to allow build for SLE11. Why forcing it on distributions that don't need it?
Not mandated by factory. Artifact of a bunch of formatter scripts that are advertised as helpful.
- replace source files by full URL I don't like this change. More precisely: I hate it.
Not mandated by factory. (But contributors will keep suggesting it to you.)
- add "-q" to %setup Why? No idea.
Not mandated by factory.
- add "V=1" to make Why? To make build warnings harder to see?
Neither V=0 nor V=1 is mandated by openSUSE:Factory. (Some contributors may suggest V=1 to you if they once ended in a position where they had to recompile your package just to verify their hypothesis that a bad compiler flag snuck in (or out).)
- break configure line into two If it needed breaking, I would do so. This one had 32 characters... Seriously?
Not mandated by factory. Talk to the guy who edited your package out-of-band / who made that funny SR to your package.
- replace ${RPM_BUILD_ROOT} with %{buildroot} The only rule I have ever seen about this was there is no rule.
1. Blame the formatting scripts you are running, explicitly or implicitly. 2. The guidelines can change over time. If you find time, please give new wiki page versions a read.
- add %{?_smp_mflags} also to make install Well, why not. But definitely not _after_ the argument.
No particular order is mandated by factory.
- %files: expand wildcards to explicit lists Some people apparently prefer package build to fail whenever a file is added or renamed. I don't force them not to do so in their packages, could they respect my preferences in mine? Apparently not.
Neither variant is mandated by factory.
- reorder tags in specfile header I'm confused. The previous order was what spec cleaner did. Wasn't it supposed to be The Right Order(TM)? And what happens next time spec cleaner service is run? Is it going to change the order back, masking real changes?
Yes of course. And they may change %rickastley to %{rickastley}. Let me repeat -- some cleaning tools suck, and you should avoid using them if they do not do what you want them to do.
Summary: lot of meaningless churn
Almost none of which are factory issues.
while I took responsibility for the package, I apparently have no say in what should its specfile look like.
If you are the Real Maintainer (socially), you get to decide. Do not let people with "maintainer" roles, and in particular, inherited prj-level "maintainer" roles, dictate you something. (Special exceptions for topical groups, like KDE:/, GNOME:/, but not the unordered hordes of devel:libraries:c_c++/, games/, science/... list is non-exhaustive.) If I may be so bold, I have been making examples of these stylistic freedoms for a long time. `osc my pkg` says something around 400, no idea what percentage is currently in factory, but it is sizable. That speaks. Your experience appears to have been an isolated incident not connected with factory, but with a prj-level "maintainer" role. Sometimes I really have to cringe at some specfiles and wonder how they got through the opensuse factory review team. I do not want to say openSUSE is the sloppiest distros of all, but claiming openSUSE has too high a specfile standard is just untrue. -- To unsubscribe, e-mail: opensuse-packaging+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-packaging+owner@opensuse.org
Jan Engelhardt wrote:
Sometimes I really have to cringe at some specfiles and wonder how they got through the opensuse factory review team. I do not want to say openSUSE is the sloppiest distros of all, but claiming openSUSE has too high a specfile standard is just untrue.
Personally two things annoying me: 1. Some reviewers insist on adding a lot of lines to .changes files claiming that there are common practices to do so. But many packages do this differently. And personally I think it's plain wrong to add all upstream changes to fit into .changes when simply updating to a new upstream release. 2. Often I just want to contribute a new upstream release to an existing package and then after submitting it I'm expected to clean up the .spec (which I did not produce) to comply to new guide-lines. Sometimes I do but sometimes I just do not have the time and give up. Ciao, Michael.
Am Dienstag, 2. Februar 2016, 08:25:20 schrieb Michael Ströder:
Jan Engelhardt wrote:
Sometimes I really have to cringe at some specfiles and wonder how they got through the opensuse factory review team. I do not want to say openSUSE is the sloppiest distros of all, but claiming openSUSE has too high a specfile standard is just untrue.
Personally two things annoying me:
1. Some reviewers insist on adding a lot of lines to .changes files claiming that there are common practices to do so. But many packages do this differently. And personally I think it's plain wrong to add all upstream changes to fit into .changes when simply updating to a new upstream release.
+1 if someone is really interested in details, reading the changelog upstream should be sufficient. Axel -- To unsubscribe, e-mail: opensuse-packaging+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-packaging+owner@opensuse.org
On Tue, Feb 02, 2016 at 08:29:15AM +0100, Axel Braun wrote:
Am Dienstag, 2. Februar 2016, 08:25:20 schrieb Michael Ströder:
Jan Engelhardt wrote:
Sometimes I really have to cringe at some specfiles and wonder how they got through the opensuse factory review team. I do not want to say openSUSE is the sloppiest distros of all, but claiming openSUSE has too high a specfile standard is just untrue.
Personally two things annoying me:
1. Some reviewers insist on adding a lot of lines to .changes files claiming that there are common practices to do so. But many packages do this differently. And personally I think it's plain wrong to add all upstream changes to fit into .changes when simply updating to a new upstream release.
+1 if someone is really interested in details, reading the changelog upstream should be sufficient.
-1 That requires a user to download the upstream package independently just because the packaging process has intentionally truncated or abridged the information - "updated from upstream" is about as useful as the "fixed some bugs" message popular on some mobile appstores. Or am I misunderstanding how an end-user is supposed to inspect/query what has changed in a pre-built package using the tools provided by the package manager? Daniel -- To unsubscribe, e-mail: opensuse-packaging+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-packaging+owner@opensuse.org
Daniel Morris wrote:
On Tue, Feb 02, 2016 at 08:29:15AM +0100, Axel Braun wrote:
Am Dienstag, 2. Februar 2016, 08:25:20 schrieb Michael Ströder:
Personally two things annoying me: 1. Some reviewers insist on adding a lot of lines to .changes files claiming that there are common practices to do so. But many packages do this differently. And personally I think it's plain wrong to add all upstream changes to fit into .changes when simply updating to a new upstream release.
+1 if someone is really interested in details, reading the changelog upstream should be sufficient.
-1 That requires a user to download the upstream package independently just because the packaging process has intentionally truncated or abridged the information - "updated from upstream" is about as useful as the "fixed some bugs" message popular on some mobile appstores.
It's not a big deal for a package foo to add the upstream changes file in /usr/share/doc/packages/foo
Or am I misunderstanding how an end-user is supposed to inspect/query what has changed in a pre-built package using the tools provided by the package manager?
Could you please elaborate on what's an "end-user"? Ciao, Michael.
On Tue, Feb 02, 2016 at 12:22:27PM +0100, Michael Ströder wrote:
Daniel Morris wrote:
On Tue, Feb 02, 2016 at 08:29:15AM +0100, Axel Braun wrote:
Am Dienstag, 2. Februar 2016, 08:25:20 schrieb Michael Ströder:
Personally two things annoying me: 1. Some reviewers insist on adding a lot of lines to .changes files claiming that there are common practices to do so. But many packages do this differently. And personally I think it's plain wrong to add all upstream changes to fit into .changes when simply updating to a new upstream release.
+1 if someone is really interested in details, reading the changelog upstream should be sufficient.
-1 That requires a user to download the upstream package independently just because the packaging process has intentionally truncated or abridged the information - "updated from upstream" is about as useful as the "fixed some bugs" message popular on some mobile appstores.
It's not a big deal for a package foo to add the upstream changes file in /usr/share/doc/packages/foo
But that only works post-install, and I'm not sure it is safe for multiple versions of foo. Out of curiosity, from 2108 rpm packages installed on this tumbleweed system, 1128 have a directory under /usr/share/docs/packages and 329 of those have a (case insensitive) file named changelog.
Or am I misunderstanding how an end-user is supposed to inspect/query what has changed in a pre-built package using the tools provided by the package manager?
Could you please elaborate on what's an "end-user"?
Great question - in the broadest terms I expect the prospective or actual consumer of a piece of software that has been pre-built and distributed in the RPM Package Manager format :) That might include a user who is happier using a graphical front-end to either inspect software on an existing system or is looking for new/newer great software. As well as those that might want to sanity check what whizo feature $BLING is likely to come, maybe with some potential breakage/interaction ahead of installation from a command line with something like 'rpm -qp --changelog foo-x.y.z+3.rpm'. Right through to the poor sysadmin firefighting in an unconnected inhospitable windowless void wondering why a twenty year convention of passing --changelog into the RPM packaging tools returns "updated from upstream" after $EXsysadmin caused some breakage on a production system that wasn't supposed to be changed. I'm sure there are other types of end-users too. Is there concensus in the LSB & RPM specs for what is the intended and/or expected behaviour? Does the map match the terrain? Daniel -- To unsubscribe, e-mail: opensuse-packaging+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-packaging+owner@opensuse.org
On Tuesday 2016-02-02 13:45, Daniel Morris wrote:
It's not a big deal for a package foo to add the upstream changes file in /usr/share/doc/packages/foo
But that only works post-install, and I'm not sure it is safe for multiple versions of foo.
Why would that only work after the installation? "%doc ChangeLog" in your .spec - done. Even works for multiple library versions If Done Right™. (lands in /usr/share/doc/packages/libnl3/ChangeLog and /usr/share/doc/packages/libnl-1_1/ChangeLog for example.) -- To unsubscribe, e-mail: opensuse-packaging+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-packaging+owner@opensuse.org
On Tue, Feb 02, 2016 at 03:14:53PM +0100, Jan Engelhardt wrote:
On Tuesday 2016-02-02 13:45, Daniel Morris wrote:
It's not a big deal for a package foo to add the upstream changes file in /usr/share/doc/packages/foo
But that only works post-install, and I'm not sure it is safe for multiple versions of foo.
Why would that only work after the installation? "%doc ChangeLog"
How would any file from foo.rpm arrive in /usr/share/doc/packages/foo before a directive to install (or update)? That would be really worrying! I was thinking just about querying. However, having poked around a few packages I now see how horribly inconsistent the whole ChangeLog handling is. In my limited sample of one system, less than 1/6th of installed packages had a corresponding upstream changelog copied to /usr/share/doc/packages/foo. Even finding a changelog, instead of Changes etc. is hit and miss. That could be simple reflection of upstream where tarball producers don't want to repeat information already held in their source, revision control, configuration management or bug-tracking processes. I guess I've been fortunate to only be sufficiently interested in the changelog output from rpm for packages that had a good level of detail rippled through, as per the simple suggestions in the "Referring to upstream ChangeLog" guidance on the twiki. As a simple end-user I'd like to say a big thank you to all the packagers that work so hard to get that balance right. Daniel -- To unsubscribe, e-mail: opensuse-packaging+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-packaging+owner@opensuse.org
Gesendet: Dienstag, 02. Februar 2016 um 12:16 Uhr Von: "Daniel Morris"
An: "Axel Braun" Cc: opensuse-packaging@opensuse.org, opensuse-factory@opensuse.org Betreff: Re: [opensuse-packaging] Re: [opensuse-factory] why there are so few packages in Factory, revisited
1. Some reviewers insist on adding a lot of lines to .changes files claiming that there are common practices to do so. But many packages do this differently. And personally I think it's plain wrong to add all upstream changes to fit into .changes when simply updating to a new upstream release.
+1 if someone is really interested in details, reading the changelog upstream should be sufficient.
-1 That requires a user to download the upstream package independently just because the packaging process has intentionally truncated or abridged the information - "updated from upstream" is about as useful as the "fixed some bugs" message popular on some mobile appstores.
Or am I misunderstanding how an end-user is supposed to inspect/query what has changed in a pre-built package using the tools provided by the package manager?
I feel so. An end user is not interested in the change log at all. A developer or packer may be interested, but mostly in case some things dont work. And in this case you have to dig deeper anyway.... Cheers Axel -- To unsubscribe, e-mail: opensuse-packaging+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-packaging+owner@opensuse.org
On Thu, 4 Feb 2016 10:20, Axel Braun
Gesendet: Dienstag, 02. Februar 2016 um 12:16 Uhr Von: "Daniel Morris"
An: "Axel Braun" Cc: opensuse-packaging@opensuse.org, opensuse-factory@opensuse.org Betreff: Re: [opensuse-packaging] Re: [opensuse-factory] why there are so few packages in Factory, revisited 1. Some reviewers insist on adding a lot of lines to .changes files claiming that there are common practices to do so. But many packages do this differently. And personally I think it's plain wrong to add all upstream changes to fit into .changes when simply updating to a new upstream release.
+1 if someone is really interested in details, reading the changelog upstream should be sufficient.
-1 That requires a user to download the upstream package independently just because the packaging process has intentionally truncated or abridged the information - "updated from upstream" is about as useful as the "fixed some bugs" message popular on some mobile appstores.
Or am I misunderstanding how an end-user is supposed to inspect/query what has changed in a pre-built package using the tools provided by the package manager?
I feel so. An end user is not interested in the change log at all. A developer or packer may be interested, but mostly in case some things dont work. And in this case you have to dig deeper anyway....
I'd like to give some contra to Axel: I am very interested in having a competent changelog, esp. in view of long-term supported packages as seen in any Enterprise Linux and also Evergreen and we will see them in Leap. Notice of fixed bugs, foremost security (with reference to CVE or similar) is very helpful in checking whether or not a package is save enough to be in use, esp if you take public available servers. Details on new features/versions, details of solved bugs, etc should be in the %doc/Changelog, just the short and curlies in the .changes file. A short: Security fix for CVE [number], details in local Changelog, or see: [upstream-changelog-url] is enough in most most cases, other packages have no competent changelog them self, so more work is needed in the package changelog. Here the direct link to the upstream changelog is most helpful to inspect the happening without prior install of the package (or downloading the whole source). There we should think of exporting most of the package changelog into a file, e.g. %doc/Changelog.SUSE and just keeping a few last changes in full in the package changelog with a reference to this file. -- At best this should be automated, saving time and work for the packager (and reducing errors). - Yamaban. -- To unsubscribe, e-mail: opensuse-packaging+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-packaging+owner@opensuse.org
On Thu, Feb 04, 2016 at 11:58:40AM +0100, Yamaban wrote: [ 8< ]
There we should think of exporting most of the package changelog into a file, e.g. %doc/Changelog.SUSE and just keeping a few last changes in full in the package changelog with a reference to this file. -- At best this should be automated, saving time and work for the packager (and reducing errors).
Why this extra work plus content duplication? The full package change log is always accessible online. If you like to enhance the interface being able to display changes before you have a package installed I suggest to enhance zypper/ libzypp to download the change log on request and display it. Cheers, Lars -- Lars Müller [ˈlaː(r)z ˈmʏlɐ] Samba Team + SUSE Labs SUSE Linux, Maxfeldstraße 5, 90409 Nürnberg, Germany
Am Donnerstag, 4. Februar 2016, 11:58:40 schrieb Yamaban:
On Thu, 4 Feb 2016 10:20, Axel Braun
wrote: Gesendet: Dienstag, 02. Februar 2016 um 12:16 Uhr Von: "Daniel Morris"
An: "Axel Braun" Cc: opensuse-packaging@opensuse.org, opensuse-factory@opensuse.org Betreff: Re: [opensuse-packaging] Re: [opensuse-factory] why there are so few packages in Factory, revisited>> 1. Some reviewers insist on adding a lot of lines to .changes files claiming that there are common practices to do so. But many packages do this differently. And personally I think it's plain wrong to add all upstream changes to fit into .changes when simply updating to a new upstream release.
+1 if someone is really interested in details, reading the changelog upstream should be sufficient.
-1 That requires a user to download the upstream package independently just because the packaging process has intentionally truncated or abridged the information - "updated from upstream" is about as useful as the "fixed some bugs" message popular on some mobile appstores.
Or am I misunderstanding how an end-user is supposed to inspect/query what has changed in a pre-built package using the tools provided by the package manager?
I feel so. An end user is not interested in the change log at all. A developer or packer may be interested, but mostly in case some things dont work. And in this case you have to dig deeper anyway....
I'd like to give some contra to Axel: I am very interested in having a competent changelog, esp. in view of long-term supported packages as seen in any Enterprise Linux and also Evergreen and we will see them in Leap.
Notice of fixed bugs, foremost security (with reference to CVE or similar) is very helpful in checking whether or not a package is save enough to be in use, esp if you take public available servers.
Let me give you an example of a package that I just upgraded. I have added the changelog to the changes file, as this was requested earlier already: - hylafax 5.5.7 * fix ntries counter to apply to pages instead of documents (5 Dec 2015) * reject jobs rejected by the proxy (18-19 Nov 2015) * add RewriteFaxName and RewriteFaxNumber jobcontrol features (14 Nov 2015) * improve Chinese translation (7 Oct 2015) * make faxsetup fix blind references in Fontmap.HylaFAX (3 Oct 2015) * use the remote time on proxy job submisisons (17 Sep 2015) * create more-secure hosts.hfaxd passwords by default (28 Aug 2015) * add admin login feature for faxstat (27 Aug 2015) * add ProxyJobTag jobcontrol feature (26 Aug 2015) * fix grevious calculation problem with Class1RestrictPoorSenders and Class1RestrictPoorDestinations (20 Aug 2015) * add application/binary MIMEConverter (17 Jul 2015) * fix DynamicConfig for Class 1 modem data format support (15 Jul 2015) * fix crash in tagline imaging due to glyph ascent (23 Jun 2015) * stop messing with the FIFO during installs and uninstalls (16 Jun 2015) * avoid conflicts with a TTY environment variable (15 Jun 2015) * fix dataTimeout esp for modems with large buffers in non-ECM (22 May 2015) * fix dataTimeout primarily affecting 7200 bps ECM sending (15 May 2015) I do not feel this is interesting for an end user. Information about security fixes may be interesting, indeed. As in many thngs, you need to find the right balance.... Axel -- To unsubscribe, e-mail: opensuse-packaging+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-packaging+owner@opensuse.org
On Tuesday 2016-02-02 08:25, Michael Ströder wrote:
Jan Engelhardt wrote:
Sometimes I really have to cringe at some specfiles and wonder how they got through the opensuse factory review team. I do not want to say openSUSE is the sloppiest distros of all, but claiming openSUSE has too high a specfile standard is just untrue.
Personally two things annoying me:
1. Some reviewers insist on adding a lot of lines to .changes files claiming that there are common practices to do so. [...] for all upstream changes to fit into .changes when simply updating to a new upstream release.
That is utterly nonsensical to do. Call those maintainers out. 1. The package maintainer may not have as much understanding what each upstream changelog entry is supposed to mean 2. The reader of rpm -q --changelog may have even lesser understanding than that 3. Avoid redundant information. Especially point #3 mandates that the full changelog with technobabble should, if so desired, provided via %files, and the rpm-changelog be short and concise. Like freshcode.club messages. Because #2.
2. Often I just want to contribute a new upstream release to an existing package and then after submitting it I'm expected to clean up the .spec (which I did not produce) to comply to new guide-lines. Sometimes I do but sometimes I just do not have the time and give up.
Indeed an issue.. -- To unsubscribe, e-mail: opensuse-packaging+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-packaging+owner@opensuse.org
Jan Engelhardt wrote:
On Tuesday 2016-02-02 08:25, Michael Ströder wrote:
Jan Engelhardt wrote:
Sometimes I really have to cringe at some specfiles and wonder how they got through the opensuse factory review team. I do not want to say openSUSE is the sloppiest distros of all, but claiming openSUSE has too high a specfile standard is just untrue.
Personally two things annoying me:
1. Some reviewers insist on adding a lot of lines to .changes files claiming that there are common practices to do so. [...] for all upstream changes to fit into .changes when simply updating to a new upstream release.
That is utterly nonsensical to do. Call those maintainers out. 1. The package maintainer may not have as much understanding what each upstream changelog entry is supposed to mean 2. The reader of rpm -q --changelog may have even lesser understanding than that 3. Avoid redundant information.
Especially point #3 mandates that the full changelog with technobabble should, if so desired, provided via %files, and the rpm-changelog be short and concise. Like freshcode.club messages. Because #2.
Exactly my point of view. Next time someone hits [Decline] I will point to your message in the list archive. Ciao, Michael.
On Tue, 2016-02-02 at 08:25 +0100, Michael Ströder wrote:
1. Some reviewers insist on adding a lot of lines to .changes files claiming that there are common practices to do so. But many packages do this differently. And personally I think it's plain wrong to add all upstream changes to fit into .changes when simply updating to a new upstream release.
The big problem is that packagers are just too lazy to UNDERSTAND and READ what is written in the release announcemnts. So the review team has to do the job for you! Not seldom there is stuff written about new features that can be enabled if this and that is added as a build dep. Or other interesting stuff. It's not about COPYING it in, but actually putting in what is important to know about the release... if you copy, that's fine, BUT: take the time to read and understand the implications of the update. A pkg update is done in 30 seconds - doing it proper and understanding what other changes need to be done, takes a bit more time (I for one use the diff of 'old' and 'new' configure.ac to update build dependencies for example... does it take time? yes. Does it improve quality? Maybe not 'obvious' when I just rely that stuff in Factory is there as I expect it to be... but if my packages are branched, they don't just fail in weird ways for missing deps - or worse: silently disable features which nobody understands why. Packaging takes brain, just as many other task does... where upstream specifies well enough the deps, even ideas like the automatic perl and ruby updaters are doable... so updating them becomes a 0s task, and STILL conforms to the packaging guidelines... so if a BOT can do it: wht can't you? Too boring? Write a bot for your packages. Cheers, Dominique -- To unsubscribe, e-mail: opensuse-packaging+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-packaging+owner@opensuse.org
On Tuesday 02 of February 2016 01:21:42 Jan Engelhardt wrote:
If you think openSUSE suffers from a bureaucracy problem, just wait until you look elsewhere. ... Fedora: ... Debian: ...
Perhaps the difference is that they don't have OBS so that setting up one's personal toolbox repositories is much more complicated for their users, giving them more incentive to get their packages into the distribution. :-) (Which doesn't mean I would like to see this option go, no way, it's too addictive.)
On Tuesday 2016-02-02 00:10, Michal Kubecek wrote:
- replace xz compressed tarball by the original gzipped one
For years I've been told by OBS [...]
That's now years away. 4 maybe? Please don't tell me you just joined SUSE.
I'm pretty sure I still got that warning less than two years ago. Anyway, I never liked the policy so I appreciate the change.
- add "V=1" to make
Why? To make build warnings harder to see?
Neither V=0 nor V=1 is mandated by openSUSE:Factory. (Some contributors may suggest V=1 to you if they once ended in a position where they had to recompile your package just to verify their hypothesis that a bad compiler flag snuck in (or out).)
In my experience, with this kind of problem, I usually have to run a local build and examine the buildroot anyway - and sometimes even chroot into it to do some tests.
- add %{?_smp_mflags} also to make install
Well, why not. But definitely not _after_ the argument.
No particular order is mandated by factory.
This is not about style, this is about POSIX compliance. Personally, I find things like "ls dirname -l" a really bad habit. It relies on a non- standard behaviour of GNU getopt (which can be disabled by setting the POSIXLY_CORECT environment variable).
Your experience appears to have been an isolated incident not connected with factory, but with a prj-level "maintainer" role.
My experience is a bit different but I only maintain a few packages so that it might just have been a bad luck. However, the pattern of people who are not actively maintaining a package accepting requests within hours is, AFAICS, disturbingly frequent. Michal Kubeček -- To unsubscribe, e-mail: opensuse-packaging+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-packaging+owner@opensuse.org
On Tuesday 2016-02-02 09:45, Michal Kubecek wrote:
On Tuesday 02 of February 2016 01:21:42 Jan Engelhardt wrote:
- add %{?_smp_mflags} also to make install
Well, why not. But definitely not _after_ the argument.
No particular order is mandated by factory.
This is not about style, this is about POSIX compliance. Personally, I find things like "ls dirname -l" a really bad habit. It relies on a non- standard behaviour of GNU getopt (which can be disabled by setting the POSIXLY_CORECT environment variable).
If anyone wants to tackle posix correctness, they would better spend their time identifying scriptlets (%prep, %setup, etc.) for bashisms and set -p /bin/bash or fix em up, because they are not really sh-compatible at times. -- To unsubscribe, e-mail: opensuse-packaging+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-packaging+owner@opensuse.org
On Tue, 02 Feb 2016 00:10:13 +0100, Michal Kubecek wrote:
I checked the changes anyway (after all, I'm bugowner so I feel responsible for the package) and I really don't know what to think:
Looking at each point, some of them seem to make sense to me. Let's see:
https://build.opensuse.org/package/rdiff/network:utilities/tcpreplay?linkrev=base&rev=5
- replace xz compressed tarball by the original gzipped one For years I've been told by OBS that I should repack gzipped tarballs to save resources. It's not true any more? OK, it allows for checking the signature so it makes sense.
Yeah, recently the original tarball is more recommended. It allows to check the validity of the included source. I thought there is even an automatic service to do that.
- make BuildRoot specification unconditional For years I've been told that it is obsolete. The only reason I keep it is to allow build for SLE11. Why forcing it on distributions that don't need it?
This is a matter of taste, IMO.
- replace source files by full URL I don't like this change. More precisely: I hate it.
It looks ugly, but it allows again the automatic source tarball check. It's good for security.
- add "-q" to %setup Why? No idea.
For reducing the many lines in the log. Mostly a matter of taste, though.
- add "V=1" to make Why? To make build warnings harder to see?
It shows the compile flags so that the build service can catch incompatible or missing compile flags (like PIE or such). So, this is safer to set for most cases.
- break configure line into two If it needed breaking, I would do so. This one had 32 characters... Seriously?
Really a matter of taste. I can imagine that breaking per word would make patching easier, but nothing more than that.
- replace ${RPM_BUILD_ROOT} with %{buildroot} The only rule I have ever seen about this was there is no rule.
Some people are too stubborn :)
- add %{?_smp_mflags} also to make install Well, why not. But definitely not _after_ the argument.
Hmm, does the order of option have influence on make's behavior?
- add LICENSE to the package OK, why not.
IIRC, this is mandatory nowadays because OBS won't modify the spec file any longer like before.
- %files: expand wildcards to explicit lists Some people apparently prefer package build to fail whenever a file is added or renamed. I don't force them not to do so in their packages, could they respect my preferences in mine? Apparently not.
Yeah, I prefer wildcards, too.
- reorder tags in specfile header I'm confused. The previous order was what spec cleaner did. Wasn't it supposed to be The Right Order(TM)? And what happens next time spec cleaner service is run? Is it going to change the order back, masking real changes?
It's annoying, indeed. Apart of some improvements in the above, the rest are a matter of taste. It should be avoidable simply by communication between you and the project maintainer. When the same problem happens again and again, it should be treated more commonly. But unless it, I don't see this as a generic problem of submitting to FACTORY... thanks, Takashi -- To unsubscribe, e-mail: opensuse-packaging+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-packaging+owner@opensuse.org
On úterý 2. února 2016 9:03:04 CET Takashi Iwai wrote:
On Tue, 02 Feb 2016 00:10:13 +0100,
Michal Kubecek wrote:
I checked the changes anyway (after all, I'm bugowner so I feel
responsible for the package) and I really don't know what to think: Looking at each point, some of them seem to make sense to me.
Let's see:
https://build.opensuse.org/package/rdiff/network:utilities/tcpreplay?lin krev=base&rev=5
- replace xz compressed tarball by the original gzipped one
For years I've been told by OBS that I should repack gzipped tarballs to save resources. It's not true any more? OK, it allows for checking the signature so it makes sense.
Yeah, recently the original tarball is more recommended. It allows to check the validity of the included source. I thought there is even an automatic service to do that.
- make BuildRoot specification unconditional
For years I've been told that it is obsolete. The only reason I keep it is to allow build for SLE11. Why forcing it on distributions that don't need it?
This is a matter of taste, IMO.
- replace source files by full URL
I don't like this change. More precisely: I hate it.
It looks ugly, but it allows again the automatic source tarball check. It's good for security.
- add "-q" to %setup
Why? No idea.
For reducing the many lines in the log. Mostly a matter of taste, though.
- add "V=1" to make
Why? To make build warnings harder to see?
It shows the compile flags so that the build service can catch incompatible or missing compile flags (like PIE or such). So, this is safer to set for most cases.
- break configure line into two
If it needed breaking, I would do so. This one had 32 characters... Seriously?
Really a matter of taste. I can imagine that breaking per word would make patching easier, but nothing more than that.
- replace ${RPM_BUILD_ROOT} with %{buildroot}
The only rule I have ever seen about this was there is no rule.
Some people are too stubborn :)
- add %{?_smp_mflags} also to make install
Well, why not. But definitely not _after_ the argument.
Hmm, does the order of option have influence on make's behavior?
- add LICENSE to the package
OK, why not.
IIRC, this is mandatory nowadays because OBS won't modify the spec file any longer like before.
- %files: expand wildcards to explicit lists
Some people apparently prefer package build to fail whenever a file is added or renamed. I don't force them not to do so in their packages, could they respect my preferences in mine? Apparently not.
Yeah, I prefer wildcards, too. To be honest I already had some bad experience with trusting wildcards - build failing instead of incorrect package being shiped is imho better way to go.
- reorder tags in specfile header
I'm confused. The previous order was what spec cleaner did. Wasn't it supposed to be The Right Order(TM)? And what happens next time spec cleaner service is run? Is it going to change the order back, masking real changes?
It's annoying, indeed. This is matter of perspective, if you are contributing to multiple packages, reviewing more requests, "uniformity" of packages makes both a bit more easy. To be honest I do not consider packaging to be place where creativity/ uniqueness in packaging style is actually desirable.
Furthermore, there has been ongoing debate in matter of rights/duties of package maintainers - it seems that some package maintainers consider themselves to be owners of said packages, while in my opinion maintainer is more like a custodian/curator of package - so all matters of personal taste are secondary to actuall functional improvements.
Apart of some improvements in the above, the rest are a matter of taste. It should be avoidable simply by communication between you and the project maintainer. When the same problem happens again and again, it should be treated more commonly. But unless it, I don't see this as a generic problem of submitting to FACTORY...
Indeed I fail to see connection between openSUSE factory/tubmleweed submision rules, packaging guidelines and amount of packages in factory. Cheers Martin
On Tue, 02 Feb 2016 09:25:55 +0100, Martin Pluskal wrote:
- %files: expand wildcards to explicit lists
Some people apparently prefer package build to fail whenever a file is added or renamed. I don't force them not to do so in their packages, could they respect my preferences in mine? Apparently not.
Yeah, I prefer wildcards, too. To be honest I already had some bad experience with trusting wildcards - build failing instead of incorrect package being shiped is imho better way to go.
I was also bitten a couple of times, too. But it's a kind of question like whether to enable -i option of rm as default or not.
- reorder tags in specfile header
I'm confused. The previous order was what spec cleaner did. Wasn't it supposed to be The Right Order(TM)? And what happens next time spec cleaner service is run? Is it going to change the order back, masking real changes?
It's annoying, indeed. This is matter of perspective, if you are contributing to multiple packages, reviewing more requests, "uniformity" of packages makes both a bit more easy. To be honest I do not consider packaging to be place where creativity/ uniqueness in packaging style is actually desirable.
Well, in this case, the previous order was the canonical one by spec-cleaner. If the order is "wrong", spec-cleaner should be fixed at first :) Takashi -- To unsubscribe, e-mail: opensuse-packaging+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-packaging+owner@opensuse.org
Dne Út 2. února 2016 10:13:55, Takashi Iwai napsal(a):
On Tue, 02 Feb 2016 09:25:55 +0100,
- %files: expand wildcards to explicit lists
Some people apparently prefer package build to fail whenever a file is added or renamed. I don't force them not to do so in their packages, could they respect my preferences in mine? Apparently not.
Yeah, I prefer wildcards, too.
To be honest I already had some bad experience with trusting wildcards - build failing instead of incorrect package being shiped is imho better way to go. I was also bitten a couple of times, too. But it's a kind of question
Martin Pluskal wrote: like whether to enable -i option of rm as default or not.
- reorder tags in specfile header
I'm confused. The previous order was what spec cleaner did. Wasn't it supposed to be The Right Order(TM)? And what happens next time spec cleaner service is run? Is it going to change the order back, masking real changes?
It's annoying, indeed.
This is matter of perspective, if you are contributing to multiple packages, reviewing more requests, "uniformity" of packages makes both a bit more easy. To be honest I do not consider packaging to be place where creativity/ uniqueness in packaging style is actually desirable.
Well, in this case, the previous order was the canonical one by spec-cleaner. If the order is "wrong", spec-cleaner should be fixed at first :)
So just to be clear here. Michal never used spec-cleaner on the package. Because if he did he would get most of the stuff he done to look like Martin did. He used the format_spec_file service that comes from obs-format_spec_file package that me and others are trying to slowly obsolete. Also format_spec_file does not change header order, just splits lines on BuildRequires and few other tweaks. If you indeed find a bug in spec-cleaner feel free to report a bug at: https://github.com/openSUSE/spec-cleaner/issues Note the bug #76 for the obsoletion. The spec-cleaner is also always updated on all supported products so it does output exactly same result no matter what distro version you start it from. Cheers Tom -- To unsubscribe, e-mail: opensuse-packaging+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-packaging+owner@opensuse.org
On Tuesday 02 of February 2016 10:20:37 Tomáš Chvátal wrote:
So just to be clear here. Michal never used spec-cleaner on the package. Because if he did he would get most of the stuff he done to look like Martin did.
He used the format_spec_file service that comes from obs-format_spec_file package that me and others are trying to slowly obsolete. Also format_spec_file does not change header order, just splits lines on BuildRequires and few other tweaks.
Let me repeat: I did run exactly the command I was asked to run by the message I got when my first submit request was declined. If I was told to run some other tool, I would have. On the other hand, if all changes from that request really come from spec-cleaner tool, it would only support my point that forcing such strict formatting policy actively discourages people from submitting their packages to Factory. Michal Kubeček -- To unsubscribe, e-mail: opensuse-packaging+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-packaging+owner@opensuse.org
On úterý 2. února 2016 10:45:51 CET Michal Kubecek wrote:
On Tuesday 02 of February 2016 10:20:37 Tomáš Chvátal wrote:
So just to be clear here. Michal never used spec-cleaner on the package. Because if he did he would get most of the stuff he done to look like Martin did.
He used the format_spec_file service that comes from obs-format_spec_file package that me and others are trying to slowly obsolete. Also format_spec_file does not change header order, just splits lines on BuildRequires and few other tweaks.
Let me repeat: I did run exactly the command I was asked to run by the message I got when my first submit request was declined. If I was told to run some other tool, I would have.
On the other hand, if all changes from that request really come from spec-cleaner tool, it would only support my point that forcing such strict formatting policy actively discourages people from submitting their packages to Factory. Is this your opinion or are you aware of others feeling same? Also note that usage of spec-cleaner is not mandatory but just adviced, and not in all cases, as it would mess up some more complex packages.
Cheers Martin
On Tuesday 02 of February 2016 10:56:36 Martin Pluskal wrote:
On úterý 2. února 2016 10:45:51 CET Michal Kubecek wrote:
On the other hand, if all changes from that request really come from spec-cleaner tool, it would only support my point that forcing such strict formatting policy actively discourages people from submitting their packages to Factory.
Is this your opinion or are you aware of others feeling same?
Not only me. Michal Kubeček -- To unsubscribe, e-mail: opensuse-packaging+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-packaging+owner@opensuse.org
On úterý 2. února 2016 11:24:35 CET Michal Kubecek wrote:
On Tuesday 02 of February 2016 10:56:36 Martin Pluskal wrote:
On úterý 2. února 2016 10:45:51 CET Michal Kubecek wrote:
On the other hand, if all changes from that request really come from spec-cleaner tool, it would only support my point that forcing such strict formatting policy actively discourages people from submitting their packages to Factory.
Is this your opinion or are you aware of others feeling same?
Not only me.
Michal Kubeček
I would ask those having same feeling as you to speak up, but I am afraid that it would changes this thread to something like "what I don't like about Factory review" or "what I don't like about openSUSE" Martin
On Tuesday 02 of February 2016 09:25:55 Martin Pluskal wrote:
This is matter of perspective, if you are contributing to multiple packages, reviewing more requests, "uniformity" of packages makes both a bit more easy. To be honest I do not consider packaging to be place where creativity/ uniqueness in packaging style is actually desirable.
OK, let's repeat the important point: the order was not a result of my creativity or uniqueness. The tag order was result of osc service localrun format_spec_file i.e. exactly the command I was told to run by the message of first declined request. (Actually, neither was the original tag order, that came from our vim template.) No creativity, no trying to be unique, exactly the opposite. And seeing that at any time I'm supposed to expect a request with some small change hidden behind a massive reshuffle, that is really annoying.
Furthermore, there has been ongoing debate in matter of rights/duties of package maintainers - it seems that some package maintainers consider themselves to be owners of said packages, while in my opinion maintainer is more like a custodian/curator of package - so all matters of personal taste are secondary to actuall functional improvements.
I do _not_ oppose "actual functional improvements". What I oppose is deliberately overriding package maintainer's personal tastes and imposing personal tastes of someone who is not actively maintaining the package (and not going to). I would even go as far as calling that disrespectful and rude. Unless I have reason to believe the package is not actively maintained, I try to respect maintainer's preferences and keep the "minimum impact" approach. I wouldn't dare to do a complete specfile rework just to make it more likeable for me. Another popular phrase around here is "doocracy: those who do, decide". Ask yourself: who (apart from upstream authors) did most work on packaging the software and making it build and work? Who did test it? Who is going to get bugs assigned? Who is supposed to fix build broken by a Factory update? Who is supposed to do version updates and make sure they build and work? Is it you? If not, please do not force your personal preferences on that package and respect the right to decide of those "who do".
Indeed I fail to see connection between openSUSE factory/tubmleweed submision rules, packaging guidelines and amount of packages in factory.
Well, that's a pity. This means all the time I spent on polishing my yesterday's mail went in vain, at least as long as you are concerned. And it shows that the problem that discourages (not only) me from submitting packages to Factory, does exist, even if, thankfully, not all project maintainers feel the same way as you. Michal Kubeček -- To unsubscribe, e-mail: opensuse-packaging+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-packaging+owner@opensuse.org
On Tuesday 02 of February 2016 09:25:55 Martin Pluskal wrote:
This is matter of perspective, if you are contributing to multiple packages, reviewing more requests, "uniformity" of packages makes both a bit more easy. To be honest I do not consider packaging to be place where creativity/ uniqueness in packaging style is actually desirable.
OK, let's repeat the important point: the order was not a result of my creativity or uniqueness. The tag order was result of
osc service localrun format_spec_file
i.e. exactly the command I was told to run by the message of first declined request. (Actually, neither was the original tag order, that came from our vim template.) No creativity, no trying to be unique, exactly the opposite. And seeing that at any time I'm supposed to expect a request with some small change hidden behind a massive reshuffle, that is really annoying.
Furthermore, there has been ongoing debate in matter of rights/duties of package maintainers - it seems that some package maintainers consider themselves to be owners of said packages, while in my opinion maintainer is more like a custodian/curator of package - so all matters of personal taste are secondary to actuall functional improvements.
I do _not_ oppose "actual functional improvements". What I oppose is deliberately overriding package maintainer's personal tastes and imposing personal tastes of someone who is not actively maintaining the package (and not going to). I would even go as far as calling that disrespectful and rude.
Unless I have reason to believe the package is not actively maintained, I try to respect maintainer's preferences and keep the "minimum impact" approach. I wouldn't dare to do a complete specfile rework just to make it more likeable for me.
Another popular phrase around here is "doocracy: those who do, decide". Ask yourself: who (apart from upstream authors) did most work on packaging the software and making it build and work? Who did test it? Who is going to get bugs assigned? Who is supposed to fix build broken by a Factory update? Who is supposed to do version updates and make sure they build and work? Is it you? If not, please do not force your personal preferences on that package and respect the right to decide of those "who do". Where did I force something to your package - please elaborate.
Indeed I fail to see connection between openSUSE factory/tubmleweed submision rules, packaging guidelines and amount of packages in factory.
Well, that's a pity. This means all the time I spent on polishing my yesterday's mail went in vain, at least as long as you are concerned. And it shows that the problem that discourages (not only) me from submitting packages to Factory, does exist, even if, thankfully, not all project maintainers feel the same way as you. I am not sure I understand what you are trying to say here - in projects where I am maintainer, I allways tried to get as many packages as possible to Factory/Tumbleweed, in cases where someone from project where I am maintaner
On úterý 2. února 2016 10:28:03 CET Michal Kubecek wrote: tried to submit something to factory, I tried to help him (and although it might surprise you, did not force my personal tastes on them), in case when discussion in sr was not enough I tried to get in contact with this maintainer - did I not do enough? I am still not sure how my personal opinions/tastes in regard of packaging, or my feeling are either symptom or cause of lack of packages in Factory. Anyway, if it either makes you feel better or will improve amount of packages in Factory, feel free to blame either me or my stupidity. Cheers Martin
On Tuesday 02 of February 2016 10:41:06 Martin Pluskal wrote:
Anyway, if it either makes you feel better or will improve amount of packages in Factory, feel free to blame either me or my stupidity.
Let me quote from my initial e-mail (I know it was long and boring but it's second paragraph): Disclaimer: the sole purpose of this e-mail is to present my view on the problem and to give responsible people something to think about. It is definitely not meant as a complaint about the particular case (which serves only as an example) or some kind of pointing fingers. If it was about you or that one particular request, I would discuss it with you directly. To simplify the problem to the core, what I find discouraging are two patterns: 1. People doing (trying to do) unnecessary changes from the "personal preferences" category in packages they are not actively maintaining (or intending to). 2. People eagerly accepting requests for packages they are not actively maintaining so that the real maintainer doesn't have chance to respond. (I can understand this if maintainer does not respond for, say, a week; but five hours on Sunday?) When these two meet for one request, the result is really frustrating. Michal Kubeček -- To unsubscribe, e-mail: opensuse-packaging+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-packaging+owner@opensuse.org
On úterý 2. února 2016 11:39:40 CET Michal Kubecek wrote:
On Tuesday 02 of February 2016 10:41:06 Martin Pluskal wrote:
Anyway, if it either makes you feel better or will improve amount of packages in Factory, feel free to blame either me or my stupidity.
Let me quote from my initial e-mail (I know it was long and boring but it's second paragraph):
Disclaimer: the sole purpose of this e-mail is to present my view on the problem and to give responsible people something to think about. It is definitely not meant as a complaint about the particular case (which serves only as an example) or some kind of pointing fingers.
If it was about you or that one particular request, I would discuss it with you directly. OK, first let me clarify that my request consisted of several changes, some which I would consider beneficial, some were in my opinion very tiny improvements. I was actually expecting this sr to be declined and to discuss changes with you, and I was surprised when it was accepted after few hours. Given that despite your disclaimer, I am not aware of any similar issues that would occur to you, and given that you decided to use changes done by me as example, it is difficult for me not to step into debate, somehow feeling that I have been pulled into something different that I was hoping to accomplish in first place.
To simplify the problem to the core, what I find discouraging are two patterns:
1. People doing (trying to do) unnecessary changes from the "personal preferences" category in packages they are not actively maintaining (or intending to). 2. People eagerly accepting requests for packages they are not actively maintaining so that the real maintainer doesn't have chance to respond. (I can understand this if maintainer does not respond for, say, a week; but five hours on Sunday?)
When these two meet for one request, the result is really frustrating. I understand your frustration, but please also try to unedrstand my suprise and perhaps dissapointment when this topic appeared on mailing lists.
Michal Kubeček
Martin
Michal Kubecek wrote:
[...] To simplify the problem to the core, what I find discouraging are two patterns:
1. People doing (trying to do) unnecessary changes from the "personal preferences" category in packages they are not actively maintaining (or intending to). 2. People eagerly accepting requests for packages they are not actively maintaining so that the real maintainer doesn't have chance to respond. (I can understand this if maintainer does not respond for, say, a week; but five hours on Sunday?)
That happens every once in a while. I've complained about that myself a few years ago too. I think the consensus has always been that project maintainers are expected to let the actual package maintainer do their job and only step in if requests start to rot. In fact project maintainers are expected to step in when that happens. I'm not sure that's written down anywhere though. In any case I had the impression that the situation got better the last years. So I doubt that your experience with this package is the common case. One problem with network:utilities could be that there are just too many people listed as maintainer so noone really feels responsible for what happens in the devel project. So it might be worth cleaning that up and agree on best practices among those who stay maintainer in the project. cu Ludwig -- (o_ Ludwig Nussel //\ V_/_ http://www.suse.com/ SUSE Linux GmbH, GF: Felix Imendörffer, Jane Smithard, Graham Norton, HRB 21284 (AG Nürnberg) -- To unsubscribe, e-mail: opensuse-packaging+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-packaging+owner@opensuse.org
On Tuesday 2016-02-02 13:16, Ludwig Nussel wrote:
One problem with network:utilities could be that there are just too many people listed as maintainer so noone really feels responsible for what happens in the devel project.
The problem is that the "maintainer" role is too broad, as I wrote some years ago too. We _need_ a distinctive role "sponsor" or so, who — at the prj level — is only permitted to accept "create" action SRs, because that is what the bulk of people in many of the herding projects are useful for. For example, I have no idea why I got maintainer in Education; it is, at best, that sponsor role. -- To unsubscribe, e-mail: opensuse-packaging+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-packaging+owner@opensuse.org
Jan Engelhardt wrote:
On Tuesday 2016-02-02 13:16, Ludwig Nussel wrote:
One problem with network:utilities could be that there are just too many people listed as maintainer so noone really feels responsible for what happens in the devel project.
The problem is that the "maintainer" role is too broad, as I wrote some years ago too. We _need_ a distinctive role "sponsor" or so, who — at the prj level — is only permitted to accept "create" action SRs, because that is what the bulk of people in many of the herding projects are useful for.
So someone with no obligations accepts the package and others listed as maintainers are then expected to sort out the potential mess. Wouldn't that be a bit unfair? :-)
For example, I have no idea why I got maintainer in Education; it is, at best, that sponsor role.
You can always remove yourself from there if you're not interested. cu Ludwig -- (o_ Ludwig Nussel //\ V_/_ http://www.suse.com/ SUSE Linux GmbH, GF: Felix Imendörffer, Jane Smithard, Graham Norton, HRB 21284 (AG Nürnberg) -- To unsubscribe, e-mail: opensuse-packaging+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-packaging+owner@opensuse.org
Hi Michal, first: in general, I agree mostly with your sentiment, and these things have kept me keeping some things out of Factory but well maintained in a "devel project" (e.g. most of the stuff in the "vdr" repo is not in Fatory and will not be until someone steps up to maintain it there who is not me) On 02.02.2016 11:39, Michal Kubecek wrote:
To simplify the problem to the core, what I find discouraging are two patterns:
2. People eagerly accepting requests for packages they are not actively maintaining so that the real maintainer doesn't have chance to respond.
As one of the people with maintainer power over Base:System and a few other "garbage can" projects, I sometime also accept submissions for packages that I'm not connected with. However, because I feel your pain, I usually only pick the ones with trivial changes -- a missing buildrequires or a changed library name for Factory or such. As soon as I cannot quickly find out what the change is doing with a look at the diff (and a "spec-cleaner" run clearly makes this impossible), I will not dare accepting the package. Well, unless it's lying around for a long time and was clearly broken before (e.g. build failure).
(I can understand this if maintainer does not respond for, say, a week; but five hours on Sunday?)
When these two meet for one request, the result is really frustrating.
I'm totally with you on that one :-) -- Stefan Seyfried "For a successful technology, reality must take precedence over public relations, for nature cannot be fooled." -- Richard Feynman -- To unsubscribe, e-mail: opensuse-packaging+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-packaging+owner@opensuse.org
Hi Michal, first: in general, I agree mostly with your sentiment, and these things have kept me keeping some things out of Factory but well maintained in a "devel project" (e.g. most of the stuff in the "vdr" repo is not in Fatory and will not be until someone steps up to maintain it there who is not me) On 02.02.2016 11:39, Michal Kubecek wrote:
To simplify the problem to the core, what I find discouraging are two patterns:
2. People eagerly accepting requests for packages they are not actively maintaining so that the real maintainer doesn't have chance to respond.
As one of the people with maintainer power over Base:System and a few other "garbage can" projects, I sometime also accept submissions for packages that I'm not connected with. However, because I feel your pain, I usually only pick the ones with trivial changes -- a missing buildrequires or a changed library name for Factory or such. As soon as I cannot quickly find out what the change is doing with a look at the diff (and a "spec-cleaner" run clearly makes this impossible), I will not dare accepting the package. Well, unless it's lying around for a long time and was clearly broken before (e.g. build failure).
(I can understand this if maintainer does not respond for, say, a week; but five hours on Sunday?)
When these two meet for one request, the result is really frustrating.
I'm totally with you on that one :-) -- Stefan Seyfried "For a successful technology, reality must take precedence over public relations, for nature cannot be fooled." -- Richard Feynman -- To unsubscribe, e-mail: opensuse-packaging+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-packaging+owner@opensuse.org
Stefan Seyfried wrote:
first: in general, I agree mostly with your sentiment, and these things have kept me keeping some things out of Factory but well maintained in a "devel project" (e.g. most of the stuff in the "vdr" repo is not in Fatory and will not be until someone steps up to maintain it there who is not me)
Huh? IIUC Michal's main concern was the feeling of losing control over his package. In the vdr project there are only two maintainers, you and me. I didn't do anything actually wrt vdr the last few years. Ie in practice there's a monarchy with you as the king of vdr land. So overly eager project maintainers can hardly serve as excuse there :-) cu Ludwig -- (o_ Ludwig Nussel //\ V_/_ http://www.suse.com/ SUSE Linux GmbH, GF: Felix Imendörffer, Jane Smithard, Graham Norton, HRB 21284 (AG Nürnberg) -- To unsubscribe, e-mail: opensuse-packaging+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-packaging+owner@opensuse.org
On 04.02.2016 09:50, Ludwig Nussel wrote:
Stefan Seyfried wrote:
first: in general, I agree mostly with your sentiment, and these things have kept me keeping some things out of Factory but well maintained in a "devel project" (e.g. most of the stuff in the "vdr" repo is not in Fatory and will not be until someone steps up to maintain it there who is not me)
Huh? IIUC Michal's main concern was the feeling of losing control over his package. In the vdr project there are only two maintainers, you and me. I didn't do anything actually wrt vdr the last few years. Ie in practice there's a monarchy with you as the king of vdr land. So overly eager project maintainers can hardly serve as excuse there :-)
No, my answer was to the question "why are these packages not in Factory". And bureaucracy is one of the reasons. I'm totally fine with your co-maintainership of the vdr repos :-) -- Stefan Seyfried "For a successful technology, reality must take precedence over public relations, for nature cannot be fooled." -- Richard Feynman -- To unsubscribe, e-mail: opensuse-packaging+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-packaging+owner@opensuse.org
On Tue, Feb 02, Michal Kubecek wrote:
- replace xz compressed tarball by the original gzipped one For years I've been told by OBS that I should repack gzipped tarballs to save resources. It's not true any more? OK, it allows for checking the signature so it makes sense.
Before anyone jumps in to "improve security": not using .xz wastes alot of (local) diskspace and download bandwidth. Olaf -- To unsubscribe, e-mail: opensuse-packaging+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-packaging+owner@opensuse.org
On Tuesday 2016-02-02 09:52, Olaf Hering wrote:
On Tue, Feb 02, Michal Kubecek wrote:
- replace xz compressed tarball by the original gzipped one For years I've been told by OBS that I should repack gzipped tarballs to save resources. It's not true any more? OK, it allows for checking the signature so it makes sense.
Before anyone jumps in to "improve security": not using .xz wastes alot of (local) diskspace and download bandwidth.
Not using brotli wastes diskspace (and time). It's kind of new, but compressed as well as xz, but in the time that it would take gzip. Try it with linux-4*.tar. And anyone who wants speed should be on LZO rather than gzip. Anyhow, you need to tell upstream those things. THEY are the ones who should be using efficient compression. THEY are the ones who should be creating .sig/.asc files for the uncompressed archive (like kernel.org). But upstream is a denomination of people which often tend to ignore all these matters, just like the symbol versioning thing. -- To unsubscribe, e-mail: opensuse-packaging+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-packaging+owner@opensuse.org
*snip* Just to reduce the length and fuzz I deleted the whole thread. After checking this in depth myself there are few facts that Michal decided politely to ignore. First and utmost is that Martin created the sr# like any other contributor which was then accepted to the project. Funny thing is that the accepter is actually also PACKAGE maintainer, not just project maintainer, so nobody violated any rule about project maintainer abusing package maintainer. Then he spents almost entire email complaining about arbitrary changes to the spec file that he didn't agree on and basically deride Martin's changes because he feels they are not same like he would do them. That is seriously ugly thing to do over public channel, and if one really feels like doing so he should first actually clarify with the author of the changes privately via email/irc/whatever rather than start a full blown flame on 2 mailinglists at once. Especialy since his co-maintainer actually approved the changes, thus he should first speak with him to figure out why/what was done. I would say now Michal should come forward and apologize for actually starting flamewar that was based on basically completely false claim. One part of Michal's complain has merit regardless of wether there was something wrong done or not. We should've had already docummented what are roles and responsibilities of packagers and project maintainers. For that I will create a wiki page that will try to explain the difference and expectations. Later on we could even get OBS to implement the rules written there with some checks so noone could by accident breakt them. But for that we would need to bribe the OBS with cookies. :-) I will write this wiki draft Today in the evening (CEST) or Tomorrow in the afternoon. Cheers Tom -- To unsubscribe, e-mail: opensuse-packaging+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-packaging+owner@opensuse.org
On Wednesday 03 of February 2016 15:03:07 Tomáš Chvátal wrote: ...
I would say now Michal should come forward and apologize for actually starting flamewar that was based on basically completely false claim. ...
Perhaps you should rather apologize for trying to publicly accuse me of things that are not true, without trying to understand what I wrote and why - and completely ignoring the even the preliminary explanation in the beginning of the first mail. But based on my previous experience, I can't say I'm surprised. Anyway, thanks for confirming the fears I had before trying to open the discussion (and which prevented me from joining the previous one) were well founded. I'm just really sad that I spent two hours trying to carefully formulate the e-mail so that everyone understands it's not meant as an attack against a particular person but a demonstration of a generic problem (problem which several people already confirmed to see) - and still some people completely missed the point. Michal Kubeček -- To unsubscribe, e-mail: opensuse-packaging+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-packaging+owner@opensuse.org
Dne St 3. února 2016 15:19:08, Michal Kubecek napsal(a):
On Wednesday 03 of February 2016 15:03:07 Tomáš Chvátal wrote: ...
I would say now Michal should come forward and apologize for actually starting flamewar that was based on basically completely false claim.
...
Perhaps you should rather apologize for trying to publicly accuse me of things that are not true, without trying to understand what I wrote and why - and completely ignoring the even the preliminary explanation in the beginning of the first mail. But based on my previous experience, I can't say I'm surprised.
Anyway, thanks for confirming the fears I had before trying to open the discussion (and which prevented me from joining the previous one) were well founded. I'm just really sad that I spent two hours trying to carefully formulate the e-mail so that everyone understands it's not meant as an attack against a particular person but a demonstration of a generic problem (problem which several people already confirmed to see) - and still some people completely missed the point.
Michal, do you seriously consider writing 1 paragraf stating "this is to improve situation" and then actually spent 15 paragrafs just bashing the guys for something that was completely correct according to all the rules as positive feedback? You used FALSE claim just to proceed with your agenda and actually most people chipped in to your train on that false assumption. Just notice Dominique actually appologizing when he realized the fact. If you would really want to change anything and being decent about it you could just write down 2 mails. 1st focusing on that you foresee the need for documented process for updates wihtout using any examples or actually thinking up one completely unrelated to the reality. 2nd to packaging if you really want to change the rules and requirements on how specs should look like, where it would be discussed like any other rule that is changed or added. Honestly tell me that you would not feel offended if somebody started with "please take this as example and not as insult" and then bashed completely for example week of your work? Tomas -- To unsubscribe, e-mail: opensuse-packaging+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-packaging+owner@opensuse.org
On Wednesday 03 of February 2016 15:33:18 Tomáš Chvátal wrote:
do you seriously consider writing 1 paragraf stating "this is to improve situation" and then actually spent 15 paragrafs just bashing the guys for something that was completely correct according to all the rules as positive feedback?
It was no bashing. I'm sorry you can't see it or don't want to see it.
You used FALSE claim just to proceed with your agenda and actually most people chipped in to your train on that false assumption. Just notice Dominique actually appologizing when he realized the fact.
No "FALSE claim". I can't read and understand mails for you. I really can't. Yes, the person who accepted the request was by mistake assigned a maintainer role. I know it (I did take a look once the request was accepted) - and I knew it when I wrote the e-mail (I might have even mentioned it once). But that doesn't really change anything. If you tried understand the problem, you might have seen it.
Honestly tell me that you would not feel offended if somebody started with "please take this as example and not as insult" and then bashed completely for example week of your work?
Hm... so I didn't just write "a request full of changes adjusting the specfile according to one's personal preferences". Instead, I tried to explain why I see it as one. And all this effort bought me an accusation of "bashing". The reality is that there are people (not only me) who find what I described a problem which can discourage people from submitting packages to Factory. And "people" is definitely not only me, I already received mails confirming other people feel the same. You can either learn from it or just wave your hand and try to shoot the messenger. Pity you chose the latter, hopefully there are others who won't. Michal Kubeček -- To unsubscribe, e-mail: opensuse-packaging+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-packaging+owner@opensuse.org
On 02/03/2016 09:03 AM, Tomáš Chvátal wrote:
*snip* <snip> One part of Michal's complain has merit regardless of wether there was something wrong done or not. We should've had already docummented what are roles and responsibilities of packagers and project maintainers. For that I will create a wiki page that will try to explain the difference and expectations.
You don't have to start from scrtch. Here is a start that I had circulated to a few people but we never got finish it. The development of openSUSE Factory and the subsequent release of the distribution is supported the technical roles outlined below. The roles definition provide a general guidline and set basic expectetions. Each person filling any one or multiple roles has the freedom to interpret these guideline to fit her or his desired participation level. Role: Package maintainer for a package in a devel project for factory: - You are generally encouraged to submit your package to factory and follow the factory maintenance process - Your package should always build in the devel project and in factory, if applicable, build failures should be addressed in a reasonable time frame, 8-10 days - You are encouraged to respond to e-mail about quesntions or concerns with your package within a reasonable time frame (5 days) - You should review and accept submit request in a reasonable amount of time, 8-10 days from the time the submit request is posted - Preferably you will have 2-4 co-maintainers for the package - You have sufficient knowledge about the package to file valid upstream bug reports that are usable by the upstream developers to fix the problem - You respond to bug reports on you package(s) within a reasonable amount of time, 8-10 days - You keep the package reasonably up to date taking into consideration the upstream release cycle, the openSUSE release cycle and the general stbility of upstream releases - You develop a general understanding of the impact of changes to your package on the distribution. For example a "broken" package in a scripting language will have less impact on the distribution than a broken compiler of C runtime environment. Role: Devel project maintainer: - You do not have to maintain any packages, but you may - You should not accept submit requests until the package maintainer in your project had a chance to review and accept the request. If after a reasonable amount of time the SR has not been reviewed or accepted you may accept the SR without review, if the package still builds - You are responsible to follow up with package maintainers for packages that show breakage for more than a reasonable amount of time (8-10 days) - You may need to communicate or facilitate communication if there are major changes in your devel project that have an impact on the distribution - You should be able to communicate clearly with the package maintainers Role: Factory Reviewer: - Review packages submissions to openSUSE:Factory, openSUSE:Factory:NonFree and maintained products based on the guidlines at http://en.opensuse.org/openSUSE:Factory_review - communicate with the submitter in case of problems Role: Factory Maintainer/Release team member: Later, Robert -- Robert Schweikert MAY THE SOURCE BE WITH YOU Public Cloud Architect LINUX rjschwei@suse.com IRC: robjo
participants (20)
-
Andreas Färber
-
Axel Braun
-
Axel Braun
-
Daniel Morris
-
Dominique Leuenberger / DimStar
-
Jan Engelhardt
-
Lars Müller
-
Ludwig Nussel
-
Marcus Meissner
-
Martin Pluskal
-
Martin Pluskal
-
Michael Ströder
-
Michal Kubecek
-
Olaf Hering
-
Philipp Thomas
-
Robert Schweikert
-
Stefan Seyfried
-
Takashi Iwai
-
Tomáš Chvátal
-
Yamaban