[opensuse-factory] status distribution
Hi, It's not exactly green and not exactly red, it's rather yellowish. At the moment we have basically two problems: getting down the build failures resulting out of glibc 2.10 updates (while automake 1.11 is already at the horizon). Reorg of factory is happening happily next to it and I hope to get the live cds building today again. At the same time we experiment with -as-needed being the default option and see quite a lot of new build failures, so I'm not sure we will see that. Let me give you some background on this: as-needed changes the way the linked handles shared libraries and is promoted mainly by gentoo devs (see http://bugs.gentoo.org/234710 and http://www.gentoo.org/proj/en/qa/asneeded.xml). As you can see in the later link, the section of benefits is _much_ shorter than the section with problems (even though a lot of the problems are gentoo specific). So there is one reason why we would want it: people say openSUSE is slow and others are faster because ..., use as-needed, ... Noone so far could explain why as-needed can be faster, but surely is a lot cleaner dependency tree. And then there are tons of build failures because packages are old and/or developers don't care too much for linking and dump libraries in random order and random number into the link line and let the linker sort it out. But sometimes a library looks unneeded and is only needed later -> boom. These cases are easy to detect and disabled by changing the binutils default back for these packages. And then there are problems that will only appear during testing and those are the most interesting: * packages have broken configure checks and will silently disable features that would be in without as-needed * C++ code in shared libraries may depend on the correct order of initializing static objects. This hit e.g. KDE years ago, as-needed became popular enough long ago to don't make this a real problem. So I would like to hear feedback on this. We have an alternative of course: leaving binutils as it is and only set SUSE_ASNEEDED=1 in packages where we expect real benefits. Or we set SUSE_ASNEEDED=0 in all packages creating a build problem and you guys say we test enough to make as-needed useful. Greetings, Stephan -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org For additional commands, e-mail: opensuse-factory+help@opensuse.org
2009/6/17 Stephan Kulow
So I would like to hear feedback on this. We have an alternative of course: leaving binutils as it is and only set SUSE_ASNEEDED=1 in packages where we expect real benefits. Or we set SUSE_ASNEEDED=0 in all packages creating a build problem and you guys say we test enough to make as-needed useful.
If there are a Iot of build fails I would make a full Factory build with SUSE_ASNEEDED=1 and create an openSUSE:Factory:asneeded project similar to openSUSE:Factory:Gcc44 with a link to the packages that failed (with the build log, not sure if that would require an additional build). If there is real interest I would expect those packages (or the library packages they depend on with the real problem) to be fixed for 11.2 release. If so, set SUSE_ASNEEDED=1 again for all post-11.2 Factory builds and we will have all the 11.3 development time to find and fix the packages that build but with problems... plus any new build fail, but I don't really expect any package build starting to fail because of --as-needed between now and 11.2 release. Meanwhile one could put a ban to SUSE_ASNEEDED=1 for packages that contain libraries (or allow it only with also "--no-allow-shlib-undefined") but allow each packager to set SUSE_ASNEEDED=1 for his packages with only executables if they want to. I would really like to have the full distro building with --as-needed, call it a personal obsession. I had some free time recently and I have submitted a few patches to fix packages that failed with --as-needed... a few because they were all I knew. Having a list of problematic packages would be really helpful. -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org For additional commands, e-mail: opensuse-factory+help@opensuse.org
Am Mittwoch 17 Juni 2009 schrieb Cristian Morales Vega:
2009/6/17 Stephan Kulow
: So I would like to hear feedback on this. We have an alternative of course: leaving binutils as it is and only set SUSE_ASNEEDED=1 in packages where we expect real benefits. Or we set SUSE_ASNEEDED=0 in all packages creating a build problem and you guys say we test enough to make as-needed useful.
If there are a Iot of build fails I would make a full Factory build with SUSE_ASNEEDED=1 and create an openSUSE:Factory:asneeded project similar to openSUSE:Factory:Gcc44 with a link to the packages that failed (with the build log, not sure if that would require an additional build). If there is real interest I would expect those packages (or the library packages they depend on with the real problem) to be fixed for 11.2 release. If so, set SUSE_ASNEEDED=1 again for all post-11.2 Factory builds and we will have all the 11.3 development time to find and fix the packages that build but with problems... plus any new build fail, but I don't really expect any package build starting to fail because of --as-needed between now and 11.2 release. Meanwhile one could put a ban to SUSE_ASNEEDED=1 for packages that contain libraries (or allow it only with also "--no-allow-shlib-undefined") but allow each packager to set SUSE_ASNEEDED=1 for his packages with only executables if they want to.
I would really like to have the full distro building with --as-needed, call it a personal obsession. I had some free time recently and I have submitted a few patches to fix packages that failed with --as-needed... a few because they were all I knew. Having a list of problematic packages would be really helpful.
My current plan is to put SUSE_ASNEEDED=0 in all packages failing atm and then leave it to the packagers to take out and fix on their own schedule. http://www.suse.de/~coolo/asneeded has the logs I greped out - as soon as suse.de syncs again. Be patient in the next minutes :) Greetings, Stephan -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org For additional commands, e-mail: opensuse-factory+help@opensuse.org
On Wed, 2009-06-17 at 14:27 +0200, Stephan Kulow wrote:
My current plan is to put SUSE_ASNEEDED=0 in all packages failing atm and then leave it to the packagers to take out and fix on their own schedule.
http://www.suse.de/~coolo/asneeded has the logs I greped out - as soon as suse.de syncs again. Be patient in the next minutes :)
Greetings, Stephan
Maybe it's a good idea filling bugs to maintainers so they can be aware of this problem. It's good to know the number of packages failing is very low. Luis (metalgod) -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org For additional commands, e-mail: opensuse-factory+help@opensuse.org
Mittwoch 17 Juni 2009 sent Luis Medinas:
On Wed, 2009-06-17 at 14:27 +0200, Stephan Kulow wrote:
My current plan is to put SUSE_ASNEEDED=0 in all packages failing atm and then leave it to the packagers to take out and fix on their own schedule.
http://www.suse.de/~coolo/asneeded has the logs I greped out - as soon as suse.de syncs again. Be patient in the next minutes :)
Greetings, Stephan
Maybe it's a good idea filling bugs to maintainers so they can be aware of this problem. I don't think it's a bug if a package doesn't work out of the box with different linker flags.
Greetings, Stephan -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org For additional commands, e-mail: opensuse-factory+help@opensuse.org
On 17/06/09 09:01, Stephan Kulow wrote:
I don't think it's a bug if a package doesn't work out of the box with different linker flags.
Why not ? most of the build failures are bugs, packages use wrong linking order... I think that qualifies as a bug. -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org For additional commands, e-mail: opensuse-factory+help@opensuse.org
On Thursday 18 June 2009 05:48:22 Cristian Rodríguez wrote:
On 17/06/09 09:01, Stephan Kulow wrote:
I don't think it's a bug if a package doesn't work out of the box with different linker flags.
Why not ? most of the build failures are bugs, packages use wrong linking order... I think that qualifies as a bug. Oh, if we talked about filing the bug upstream I would agree. But for openSUSE, I would rather have us fixing real bugs.
As I said: I plan to disable as needed for packages that fail with it and be done. Feel free to spend your time enabling it again for packages you know to benefit from it. Greetings, Stephan -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org For additional commands, e-mail: opensuse-factory+help@opensuse.org
On Wed, Jun 17, 2009 at 03:01:52PM +0200, Stephan Kulow wrote:
Maybe it's a good idea filling bugs to maintainers so they can be aware of this problem. I don't think it's a bug if a package doesn't work out of the box with different linker flags.
I think the opposite way. Being involved in Wireshark development I
have to go and check all the big distros for changes made, find out
why and by whom and then apply what appears to be useful.
<rant>
It's *really* annoying! Why can't the package maintainers open a bug
with Wireshark for every patch that isn't 100% distro specific and
let the project maintainers decide what is acceptable and what isn't?
While I'm ranting anyway: Each and every patchfile that is part of
a source package should contain a description and information on the
author(s) - how am I supposed to give credit correctly otherwise when
I decide that a patch is useful? Normally figuring out the credits
part is more work than finding out whether to apply a patch or not.
</rant>
Ciao
Joerg
--
Joerg Mayer
2009/6/18 Joerg Mayer
On Wed, Jun 17, 2009 at 03:01:52PM +0200, Stephan Kulow wrote:
Maybe it's a good idea filling bugs to maintainers so they can be aware of this problem. I don't think it's a bug if a package doesn't work out of the box with different linker flags.
I think the opposite way. Being involved in Wireshark development I have to go and check all the big distros for changes made, find out why and by whom and then apply what appears to be useful. <rant> It's *really* annoying! Why can't the package maintainers open a bug with Wireshark for every patch that isn't 100% distro specific and let the project maintainers decide what is acceptable and what isn't?
Seriously? Because I think "now I will have to open an account in its bug tracker and...". When I'm not busy I report upstream, but when I have anything else to do the last thing I want is open yet another account in yet another bug tracker. My fault, but... perhaps OpenID will make people report more to upstream.
While I'm ranting anyway: Each and every patchfile that is part of a source package should contain a description and information on the author(s) - how am I supposed to give credit correctly otherwise when I decide that a patch is useful? Normally figuring out the credits part is more work than finding out whether to apply a patch or not. </rant>
I don't really mind about getting credit for some minor patches. But I suppose it's important to upstream because of legalities. Anyway, Wireshark 1.2.0 builds with --as-needed because of the new behavior in binutils, but it still has a problem that triggers if you use pre-2.20 binutils. The patch is available at https://build.opensuse.org/package/view_file?file=wireshark-1.2.0-asneeded.patch&package=wireshark&project=openSUSE%3AFactory and I'm the author. -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org For additional commands, e-mail: opensuse-factory+help@opensuse.org
On Thu, Jun 18, 2009 at 06:39:27PM +0200, Cristian Morales Vega wrote:
2009/6/18 Joerg Mayer
: On Wed, Jun 17, 2009 at 03:01:52PM +0200, Stephan Kulow wrote:
Maybe it's a good idea filling bugs to maintainers so they can be aware of this problem. I don't think it's a bug if a package doesn't work out of the box with different linker flags.
I think the opposite way. Being involved in Wireshark development I have to go and check all the big distros for changes made, find out why and by whom and then apply what appears to be useful. <rant> It's *really* annoying! Why can't the package maintainers open a bug with Wireshark for every patch that isn't 100% distro specific and let the project maintainers decide what is acceptable and what isn't?
Seriously? Because I think "now I will have to open an account in its bug tracker and...". When I'm not busy I report upstream, but when I have anything else to do the last thing I want is open yet another account in yet another bug tracker.
OH, now that's an interesting approach: You need to track upstream, so you have to spend quite a bit of time following that stuff, e.g. subscribe to the relevant mailing lists, find out the svn address, but you cannot be bothered to take the one time effort (per project that you maintain) to open an account?
My fault, but... perhaps OpenID will make people report more to upstream.
While I'm ranting anyway: Each and every patchfile that is part of a source package should contain a description and information on the author(s) - how am I supposed to give credit correctly otherwise when I decide that a patch is useful? Normally figuring out the credits part is more work than finding out whether to apply a patch or not. </rant>
I don't really mind about getting credit for some minor patches. But I suppose it's important to upstream because of legalities. Anyway, Wireshark 1.2.0 builds with --as-needed because of the new behavior in binutils, but it still has a problem that triggers if you use pre-2.20 binutils. The patch is available at https://build.opensuse.org/package/view_file?file=wireshark-1.2.0-asneeded.patch&package=wireshark&project=openSUSE%3AFactory and I'm the author.
Now you're really funny: In order to access that URL I need to create
an account .....
But what I was really complaining about is that there doesn't seem to be
any requirement to put the necessary information directly into the patches.
Ciao
Joerg
--
Joerg Mayer
Le vendredi 19 juin 2009, à 00:41 +0200, Joerg Mayer a écrit :
On Thu, Jun 18, 2009 at 06:39:27PM +0200, Cristian Morales Vega wrote:
use pre-2.20 binutils. The patch is available at https://build.opensuse.org/package/view_file?file=wireshark-1.2.0-asneeded.patch&package=wireshark&project=openSUSE%3AFactory and I'm the author.
Now you're really funny: In order to access that URL I need to create an account .....
You can use http://tmp.vuntz.net/opensuse-packages/browse.py?project=openSUSE:Factory&package=wireshark if you don't have an account. Vincent -- Les gens heureux ne sont pas pressés. -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org For additional commands, e-mail: opensuse-factory+help@opensuse.org
2009/6/19 Joerg Mayer
On Thu, Jun 18, 2009 at 06:39:27PM +0200, Cristian Morales Vega wrote:
2009/6/18 Joerg Mayer
: On Wed, Jun 17, 2009 at 03:01:52PM +0200, Stephan Kulow wrote:
Maybe it's a good idea filling bugs to maintainers so they can be aware of this problem. I don't think it's a bug if a package doesn't work out of the box with different linker flags.
I think the opposite way. Being involved in Wireshark development I have to go and check all the big distros for changes made, find out why and by whom and then apply what appears to be useful. <rant> It's *really* annoying! Why can't the package maintainers open a bug with Wireshark for every patch that isn't 100% distro specific and let the project maintainers decide what is acceptable and what isn't?
Seriously? Because I think "now I will have to open an account in its bug tracker and...". When I'm not busy I report upstream, but when I have anything else to do the last thing I want is open yet another account in yet another bug tracker.
OH, now that's an interesting approach: You need to track upstream, so you have to spend quite a bit of time following that stuff, e.g. subscribe to the relevant mailing lists, find out the svn address, but you cannot be bothered to take the one time effort (per project that you maintain) to open an account?
I don't mantain many packages, I'm more the random patch guy. But as I already said "my fault"... I was giving an explanation, not a justification. Why the mantainer doesn't submits upstream the patches I submit to him? Well, he probably thinks I already did.
My fault, but... perhaps OpenID will make people report more to upstream.
While I'm ranting anyway: Each and every patchfile that is part of a source package should contain a description and information on the author(s) - how am I supposed to give credit correctly otherwise when I decide that a patch is useful? Normally figuring out the credits part is more work than finding out whether to apply a patch or not. </rant>
I don't really mind about getting credit for some minor patches. But I suppose it's important to upstream because of legalities. Anyway, Wireshark 1.2.0 builds with --as-needed because of the new behavior in binutils, but it still has a problem that triggers if you use pre-2.20 binutils. The patch is available at https://build.opensuse.org/package/view_file?file=wireshark-1.2.0-asneeded.patch&package=wireshark&project=openSUSE%3AFactory and I'm the author.
Now you're really funny: In order to access that URL I need to create an account .....
But what I was really complaining about is that there doesn't seem to be any requirement to put the necessary information directly into the patches.
There is: http://en.opensuse.org/SUSE_Package_Conventions/RPM_Style#1.15._Patch_Tag, but... I think I have NEVER found a name in an openSUSE patch. Few times an explanation. There is also http://en.opensuse.org/GNOME/Patches, that I don't know why it's under GNOME. But in general (perhaps it's better for Gnome packages) isn't very respected. ...someone feels like creating some rpmlint checks? -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org For additional commands, e-mail: opensuse-factory+help@opensuse.org
Le vendredi 19 juin 2009, à 01:46 +0200, Cristian Morales Vega a écrit :
There is also http://en.opensuse.org/GNOME/Patches, that I don't know why it's under GNOME. But in general (perhaps it's better for Gnome packages) isn't very respected.
It got promoted to http://en.opensuse.org/Packaging/Patches a few days ago -- that's part of our packaging guidelines, now. Vincent -- Les gens heureux ne sont pas pressés. -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org For additional commands, e-mail: opensuse-factory+help@opensuse.org
On Fri, 2009-06-19 at 01:46 +0200, Cristian Morales Vega wrote: [...]
There is also http://en.opensuse.org/GNOME/Patches, that I don't know why it's under GNOME. But in general (perhaps it's better for Gnome packages) isn't very respected.
...someone feels like creating some rpmlint checks?
The GNOME team uses these tags to keep track of what patches etc we have. The result of properly tagging our patches can be seen at http://tmp.vuntz.net/opensuse-packages/patch.py We also do rpmlint checks on our packages, which can be found at http://tmp.vuntz.net/opensuse-packages/rpmlint.py Cheers, Magnus -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org For additional commands, e-mail: opensuse-factory+help@opensuse.org
On 6/17/2009 at 14:27, Stephan Kulow
wrote: My current plan is to put SUSE_ASNEEDED=0 in all packages failing atm and then leave it to the packagers to take out and fix on their own schedule. http://www.suse.de/~coolo/asneeded has the logs I greped out - as soon as suse.de syncs again. Be patient in the next minutes :)
You have libproxy on your list. SVN commit 303 addresses this. It will be part of release 0.3, which I try to push out before feature freeze of openSUSE 11.2 :) Dominique -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org For additional commands, e-mail: opensuse-factory+help@opensuse.org
2009/6/17 Stephan Kulow
Am Mittwoch 17 Juni 2009 schrieb Cristian Morales Vega:
2009/6/17 Stephan Kulow
: So I would like to hear feedback on this. We have an alternative of course: leaving binutils as it is and only set SUSE_ASNEEDED=1 in packages where we expect real benefits. Or we set SUSE_ASNEEDED=0 in all packages creating a build problem and you guys say we test enough to make as-needed useful.
If there are a Iot of build fails I would make a full Factory build with SUSE_ASNEEDED=1 and create an openSUSE:Factory:asneeded project similar to openSUSE:Factory:Gcc44 with a link to the packages that failed (with the build log, not sure if that would require an additional build). If there is real interest I would expect those packages (or the library packages they depend on with the real problem) to be fixed for 11.2 release. If so, set SUSE_ASNEEDED=1 again for all post-11.2 Factory builds and we will have all the 11.3 development time to find and fix the packages that build but with problems... plus any new build fail, but I don't really expect any package build starting to fail because of --as-needed between now and 11.2 release. Meanwhile one could put a ban to SUSE_ASNEEDED=1 for packages that contain libraries (or allow it only with also "--no-allow-shlib-undefined") but allow each packager to set SUSE_ASNEEDED=1 for his packages with only executables if they want to.
I would really like to have the full distro building with --as-needed, call it a personal obsession. I had some free time recently and I have submitted a few patches to fix packages that failed with --as-needed... a few because they were all I knew. Having a list of problematic packages would be really helpful.
My current plan is to put SUSE_ASNEEDED=0 in all packages failing atm and then leave it to the packagers to take out and fix on their own schedule.
Between, where SUSE_ASNEEDED should be setted? Not really sure about what I can put in the package/project metadata in the BS and how should be setted ("export SUSE_ASNEEDED=1"?).
http://www.suse.de/~coolo/asneeded has the logs I greped out - as soon as suse.de syncs again. Be patient in the next minutes :)
MozillaThunderbird buildlog 16,4MiB?? Ok...will be the last one ;-) -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org For additional commands, e-mail: opensuse-factory+help@opensuse.org
Stephan Kulow wrote:
My current plan is to put SUSE_ASNEEDED=0 in all packages failing atm and then leave it to the packagers to take out and fix on their own schedule.
http://www.suse.de/~coolo/asneeded has the logs I greped out - as soon as suse.de syncs again. Be patient in the next minutes :)
/me giggles. Could have sworn our palette of Mozilla apps (MozillaFirefox, MozillaThunderbird, seamonkey) was on that list - and I'm right! :P I wonder though if we should move to Thunderbird 3.0 Beta and SeaMonkey 2.0 Alpha/Beta builds for Factory though, from all we hear those are not really less stable than the Thunderbird 2 and SeaMonkey 1.1 "stable" versions, but catch up with Firefox 3.5 code-wise (could help with this -as-needed stuff perhaps) and both should release finals before openSUSE 11.2 goes gold. Wolfgang as OBS package manager for Mozilla builds may have some input into that, I guess. Robert Kaiser -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org For additional commands, e-mail: opensuse-factory+help@opensuse.org
Robert Kaiser schrieb:
/me giggles.
:-(
Could have sworn our palette of Mozilla apps (MozillaFirefox, MozillaThunderbird, seamonkey) was on that list - and I'm right! :P
I wonder though if we should move to Thunderbird 3.0 Beta and SeaMonkey 2.0 Alpha/Beta builds for Factory though, from all we hear those are not really less stable than the Thunderbird 2 and SeaMonkey 1.1 "stable" versions, but catch up with Firefox 3.5 code-wise (could help with this -as-needed stuff perhaps) and both should release finals before openSUSE 11.2 goes gold.
Wolfgang as OBS package manager for Mozilla builds may have some input into that, I guess.
Yes ;-) Actually I'm waiting impatiently for a Thunderbird 3.0b3 release to put it into Factory. Putting 3.0b2 there would just be too old and a nightly breaks far too many locales to have it in Factory IMHO. Almost the same for SeaMonkey while there is the difference that I didn't care yet to add a locale package anyway so updating to a current nightly would be possible. That's more a decision I'd like to proxy to you Robert as SeaMonkey project lead. If you'd like to see SeaMonkey 2.0a/b earlier than later in Factory I'm fine with it. Just let me know. Wolfgang -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org For additional commands, e-mail: opensuse-factory+help@opensuse.org
Wolfgang Rosenauer wrote:
Actually I'm waiting impatiently for a Thunderbird 3.0b3 release to put it into Factory. Putting 3.0b2 there would just be too old and a nightly breaks far too many locales to have it in Factory IMHO.
Ah, Ok, sounds reasonable. All the pushing out of that beta didn't really help there, I guess :( Thunderbird people tell me that they have the major patch that blocked the beta in now and should get the beta done in the next weeks (SeaMonkey will go Beta 1 roughly in sync, BTW).
Almost the same for SeaMonkey while there is the difference that I didn't care yet to add a locale package anyway so updating to a current nightly would be possible. That's more a decision I'd like to proxy to you Robert as SeaMonkey project lead. If you'd like to see SeaMonkey 2.0a/b earlier than later in Factory I'm fine with it. Just let me know.
SeaMonkey 2 locale packages only started to be able to reasonably work in the last weeks, so good that you didn't look into that earlier. ;-) That said, I think that current SeaMonkey 2.0b1pre nightlies are almost as stable as 1.1.x but significantly better not only in functionality but also is compatibility with today's operating systems. I think it's definitely worth to put that code into Factory. Robert Kaiser -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org For additional commands, e-mail: opensuse-factory+help@opensuse.org
On Wednesday 17 June 2009 22:55:57 Robert Kaiser wrote:
That said, I think that current SeaMonkey 2.0b1pre nightlies are almost as stable as 1.1.x but significantly better not only in functionality but also is compatibility with today's operating systems. I think it's definitely worth to put that code into Factory.
Will we see releases of SeaMonkey and Thunderbird in time for 11.2? In that case, I agree with moving forward, Andreas -- Andreas Jaeger, Director Platform / openSUSE, aj@suse.de SUSE LINUX Products GmbH, GF: Markus Rex, HRB 16746 (AG Nürnberg) Maxfeldstr. 5, 90409 Nürnberg, Germany GPG fingerprint = 93A3 365E CE47 B889 DF7F FED1 389A 563C C272 A126
Andreas Jaeger schrieb:
On Wednesday 17 June 2009 22:55:57 Robert Kaiser wrote:
That said, I think that current SeaMonkey 2.0b1pre nightlies are almost as stable as 1.1.x but significantly better not only in functionality but also is compatibility with today's operating systems. I think it's definitely worth to put that code into Factory.
Will we see releases of SeaMonkey and Thunderbird in time for 11.2? In that case, I agree with moving forward,
According to http://en.opensuse.org/Roadmap/11.2 we are talking about: Sun, Aug 02 openSUSE 11.2 Milestone 5 Release as "Feature and version freeze for the complete distribution". I'm not quite sure if we will have final releases at that point but it could still make sense to ship what is the latest beta then. Wolfgang -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org For additional commands, e-mail: opensuse-factory+help@opensuse.org
On Thursday 18 June 2009 10:34:08 Wolfgang Rosenauer wrote:
Andreas Jaeger schrieb:
On Wednesday 17 June 2009 22:55:57 Robert Kaiser wrote:
That said, I think that current SeaMonkey 2.0b1pre nightlies are almost as stable as 1.1.x but significantly better not only in functionality but also is compatibility with today's operating systems. I think it's definitely worth to put that code into Factory.
Will we see releases of SeaMonkey and Thunderbird in time for 11.2? In that case, I agree with moving forward,
According to http://en.opensuse.org/Roadmap/11.2 we are talking about: Sun, Aug 02 openSUSE 11.2 Milestone 5 Release as "Feature and version freeze for the complete distribution".
"Patch level update of leaf-packages" applies for sure for SeaMonkey (I doubt it's on the media) and might as well for Thunderbird - which would make a deadline of 18th of August.
I'm not quite sure if we will have final releases at that point but it could still make sense to ship what is the latest beta then.
So, what is the planned final release? And then let's discuss with Coolo, Andreas -- Andreas Jaeger, Director Platform / openSUSE, aj@suse.de SUSE LINUX Products GmbH, GF: Markus Rex, HRB 16746 (AG Nürnberg) Maxfeldstr. 5, 90409 Nürnberg, Germany GPG fingerprint = 93A3 365E CE47 B889 DF7F FED1 389A 563C C272 A126
Andreas Jaeger wrote:
On Wednesday 17 June 2009 22:55:57 Robert Kaiser wrote:
That said, I think that current SeaMonkey 2.0b1pre nightlies are almost as stable as 1.1.x but significantly better not only in functionality but also is compatibility with today's operating systems. I think it's definitely worth to put that code into Factory.
Will we see releases of SeaMonkey and Thunderbird in time for 11.2? In that case, I agree with moving forward,
Two months ago I would have said "we'll surely make August" for both. With the recent pushing out of things I'm not completely sure. As we very probably will stop maintaining Thunderbird 2.x and SeaMonkey 1.x well before the maintenance period of 11.2 ends, and given the stability that testers of development builds are seeing in current Thunderbird 3.0 and SeaMonkey 2.0 pre-releases I think it would even be better to ship betas of them with 11.2 than so-called "stable" versions of the old series. It's unfortunate that we don't have finals of the new versions right now, then this would be very much easier, that's for sure. Robert Kaiser -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org For additional commands, e-mail: opensuse-factory+help@opensuse.org
On Thursday 18 June 2009 13:32:09 Robert Kaiser wrote:
Andreas Jaeger wrote:
On Wednesday 17 June 2009 22:55:57 Robert Kaiser wrote:
That said, I think that current SeaMonkey 2.0b1pre nightlies are almost as stable as 1.1.x but significantly better not only in functionality but also is compatibility with today's operating systems. I think it's definitely worth to put that code into Factory.
Will we see releases of SeaMonkey and Thunderbird in time for 11.2? In that case, I agree with moving forward,
Two months ago I would have said "we'll surely make August" for both. With the recent pushing out of things I'm not completely sure. As we very probably will stop maintaining Thunderbird 2.x and SeaMonkey 1.x well before the maintenance period of 11.2 ends, and given the stability that testers of development builds are seeing in current Thunderbird 3.0 and SeaMonkey 2.0 pre-releases I think it would even be better to ship betas of them with 11.2 than so-called "stable" versions of the old series. It's unfortunate that we don't have finals of the new versions right now, then this would be very much easier, that's for sure.
Coolo's call in this case - and I support the proposal to go to the new versions and ship them even as betas if needed (and consider updating them via online update after release). But this is quite an exception! Andreas -- Andreas Jaeger, Director Platform / openSUSE, aj@suse.de SUSE LINUX Products GmbH, GF: Markus Rex, HRB 16746 (AG Nürnberg) Maxfeldstr. 5, 90409 Nürnberg, Germany GPG fingerprint = 93A3 365E CE47 B889 DF7F FED1 389A 563C C272 A126
Am Donnerstag 18 Juni 2009 schrieb Andreas Jaeger:
Coolo's call in this case - and I support the proposal to go to the new versions and ship them even as betas if needed (and consider updating them via online update after release). But this is quite an exception!
Exception on what base? Sorry, rules are rules and the 11.2 window really is large enough not to need exceptions for the case of it. And of course development versions have less testers, so there will always be less bugs in the development version than in the released version. Greetings, Stephan -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org For additional commands, e-mail: opensuse-factory+help@opensuse.org
Stephan Kulow schrieb:
Am Donnerstag 18 Juni 2009 schrieb Andreas Jaeger:
Coolo's call in this case - and I support the proposal to go to the new versions and ship them even as betas if needed (and consider updating them via online update after release). But this is quite an exception!
Exception on what base? Sorry, rules are rules and the 11.2 window really is large enough not to need exceptions for the case of it. And of course development versions have less testers, so there will always be less bugs in the development version than in the released version.
Ok, I don't care too much as the need for having to rely on openSUSE "base" is declining. Anyway could you please point me to the rules you are referring to? So to make that clear one more time we are talking about updating MozillaThunderbird and seamonkey to beta versions in worst case before openSUSE will enter feature freeze. Either we do that and allow beta versions what we did pretty often in the past (especially by introducing alpha quality package management which obviously wasn't named alpha or beta) or the other alternative would be to drop MozillaThunderbird and seamonkey from Factory _now_ as I'm pretty sure that security support will be completely stopped once Thunderbird 3 and seamonkey 2 got released as it's already only limited support since Firefox 2 went out of maintenance upstream. I would choose the former but if that's not possible I would choose the latter (if I were you or your security team). Wolfgang -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org For additional commands, e-mail: opensuse-factory+help@opensuse.org
Am Donnerstag 18 Juni 2009 schrieb Wolfgang Rosenauer:
Stephan Kulow schrieb:
Am Donnerstag 18 Juni 2009 schrieb Andreas Jaeger:
Coolo's call in this case - and I support the proposal to go to the new versions and ship them even as betas if needed (and consider updating them via online update after release). But this is quite an exception!
Exception on what base? Sorry, rules are rules and the 11.2 window really is large enough not to need exceptions for the case of it. And of course development versions have less testers, so there will always be less bugs in the development version than in the released version.
Ok, I don't care too much as the need for having to rely on openSUSE "base" is declining.
Anyway could you please point me to the rules you are referring to?
So to make that clear one more time we are talking about updating MozillaThunderbird and seamonkey to beta versions in worst case before openSUSE will enter feature freeze. Either we do that and allow beta versions what we did pretty often in the past (especially by introducing alpha quality package management which obviously wasn't named alpha or beta) or the other alternative would be to drop MozillaThunderbird and seamonkey from Factory _now_ as I'm pretty sure that security support will be completely stopped once Thunderbird 3 and seamonkey 2 got released as it's already only limited support since Firefox 2 went out of maintenance upstream. I would choose the former but if that's not possible I would choose the latter (if I were you or your security team).
The rules are about version updates and we need to be aware on what base we grant exceptions because "we can discuss it" is not a rule - we have way too many packages to make it feasible to discuss with me about updating them. And this is basically what I want to avoid: having rules clearly expressing the cases and as you say yourself: the openSUSE release is just a base, still many people don't do updates, so I don't want to ship beta versions if there aren't really, really good reasons. And not having security updates for a package that surely depends on it, is a good reason. Case closed and discussion archived hereby :) Greetings, Stephan -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org For additional commands, e-mail: opensuse-factory+help@opensuse.org
On Thursday 18 June 2009 14:03:06 Stephan Kulow wrote:
Am Donnerstag 18 Juni 2009 schrieb Andreas Jaeger:
Coolo's call in this case - and I support the proposal to go to the new versions and ship them even as betas if needed (and consider updating them via online update after release). But this is quite an exception!
Exception on what base? Sorry, rules are rules and the 11.2 window really is large enough not to need exceptions for the case of it. And of course development versions have less testers, so there will always be less bugs in the development version than in the released version.
I would ship these as betas based on the comments that Robert made regarding stability. We can either freeze in August - or later. Or we freeze in August and update at some time to the final version. Andreas -- Andreas Jaeger, Director Platform / openSUSE, aj@suse.de SUSE LINUX Products GmbH, GF: Markus Rex, HRB 16746 (AG Nürnberg) Maxfeldstr. 5, 90409 Nürnberg, Germany GPG fingerprint = 93A3 365E CE47 B889 DF7F FED1 389A 563C C272 A126
On Wed, 17 Jun 2009, Robert Kaiser wrote:
SeaMonkey 2 locale packages only started to be able to reasonably work in the last weeks, so good that you didn't look into that earlier. ;-)
That said, I think that current SeaMonkey 2.0b1pre nightlies are almost as stable as 1.1.x but significantly better not only in functionality but also is compatibility with today's operating systems. I think it's definitely worth to put that code into Factory.
Speaking as a SeaMonkey user, I've been using the nightlies on my work WinXP Vm for some months now and it has been working fine. Seeing this in factory would be a good thing, in my opinion. -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org For additional commands, e-mail: opensuse-factory+help@opensuse.org
On Wednesday 17 June 2009 07:27:03 am Stephan Kulow wrote:
Am Mittwoch 17 Juni 2009 schrieb Cristian Morales Vega:
2009/6/17 Stephan Kulow
: So I would like to hear feedback on this. We have an alternative of course: leaving binutils as it is and only set SUSE_ASNEEDED=1 in packages where we expect real benefits. Or we set SUSE_ASNEEDED=0 in all packages creating a build problem and you guys say we test enough to make as-needed useful.
If there are a Iot of build fails I would make a full Factory build with SUSE_ASNEEDED=1 and create an openSUSE:Factory:asneeded project similar to openSUSE:Factory:Gcc44 with a link to the packages that failed (with the build log, not sure if that would require an additional build). If there is real interest I would expect those packages (or the library packages they depend on with the real problem) to be fixed for 11.2 release. If so, set SUSE_ASNEEDED=1 again for all post-11.2 Factory builds and we will have all the 11.3 development time to find and fix the packages that build but with problems... plus any new build fail, but I don't really expect any package build starting to fail because of --as-needed between now and 11.2 release. Meanwhile one could put a ban to SUSE_ASNEEDED=1 for packages that contain libraries (or allow it only with also "--no-allow-shlib-undefined") but allow each packager to set SUSE_ASNEEDED=1 for his packages with only executables if they want to.
I would really like to have the full distro building with --as-needed, call it a personal obsession. I had some free time recently and I have submitted a few patches to fix packages that failed with --as-needed... a few because they were all I knew. Having a list of problematic packages would be really helpful.
My current plan is to put SUSE_ASNEEDED=0 in all packages failing atm and then leave it to the packagers to take out and fix on their own schedule.
http://www.suse.de/~coolo/asneeded has the logs I greped out - as soon as suse.de syncs again. Be patient in the next minutes :)
Greetings, Stephan
Not that I have a lot of input on build intricacies, but I whole heartedly agree with Stephan here. Time needs to be allowed for the development of 11.2 to "Settle" in the short amount of time between now and release so a clean up of the bugs in the packages can take place without introducing new ones with asneeded. Regardless of how much manpower is dedicated to rushing-pushing-tinkering and changing every package to build with asneeded between now and 11.2 release, reality says you will be chasing asneeded bugs all the way up to the release date and then bandaiding packages at the last minute to press DVD's and inevitably then many unintended consequences of the asneeded rush won't show up until after the DVD's are pressed because the state of build process wasn't frozen (asneeded=0/asneeded=1) long enough before release to allow adequate testing to minimize the number of "Oopses" in the final product. If adequate time isn't allowed for simply ironing out the packages with their current build procedure, I think you are just setting yourselves up for failure and being caught short on time to get it all done. Personally, I would like to see 11.2 be a clean release without 1.2G of package updates within the first 30 days of release needed to fix problems that would have been avoided if a smarter and more conservative approach to getting 11.2 DVD ready had been taken. There is only so much that can be done. The decision to freeze/decide upon the build process for packages seems like something that needs to happen yesterday so the cleanup can begin. Just my .02 that applies to any large undertaking whether writing code or changing APU turbine speed sensors on the shuttle. (there are 3 per turbine if you are interested)(Manufactured by Sunstrand) -- David C. Rankin, J.D.,P.E. Rankin Law Firm, PLLC 510 Ochiltree Street Nacogdoches, Texas 75961 Telephone: (936) 715-9333 Facsimile: (936) 715-9339 www.rankinlawfirm.com -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org For additional commands, e-mail: opensuse-factory+help@opensuse.org
2009/6/20 David C. Rankin
Not that I have a lot of input on build intricacies, but I whole heartedly agree with Stephan here. Time needs to be allowed for the development of 11.2 to "Settle" in the short amount of time between now and release so a clean up of the bugs in the packages can take place without introducing new ones with asneeded.
Regardless of how much manpower is dedicated to rushing-pushing-tinkering and changing every package to build with asneeded between now and 11.2 release, reality says you will be chasing asneeded bugs all the way up to the release date and then bandaiding packages at the last minute to press DVD's and inevitably then many unintended consequences of the asneeded rush won't show up until after the DVD's are pressed because the state of build process wasn't frozen (asneeded=0/asneeded=1) long enough before release to allow adequate testing to minimize the number of "Oopses" in the final product.
Your argument is valid but simplistic. How much is "enough time"? 11.2 will be released in 4 month and three weeks (~5months), most distros release every six months... with their release schedule we would have just missed one month. We all agree that "enough time" is needed to test changes... but nobody agrees about how much is "enough time". That's why we have a project manager... to bother him about changes we want even if there is no time and blame him about other changes that got done without time ;-) -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org For additional commands, e-mail: opensuse-factory+help@opensuse.org
On Saturday 20 June 2009 04:41:52 am Cristian Morales Vega wrote:
We all agree that "enough time" is needed to test changes... but nobody agrees about how much is "enough time". That's why we have a project manager... to bother him about changes we want even if there is no time and blame him about other changes that got done without time ;-)
IMHO, we should be conservative with this release to avoid experience with few previous. There is no better marketing than good release. -- Regards, Rajko http://news.opensuse.org/category/people-of-opensuse/ -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org For additional commands, e-mail: opensuse-factory+help@opensuse.org
Rajko M. wrote:
On Saturday 20 June 2009 04:41:52 am Cristian Morales Vega wrote:
We all agree that "enough time" is needed to test changes... but nobody agrees about how much is "enough time". That's why we have a project manager... to bother him about changes we want even if there is no time and blame him about other changes that got done without time ;-)
IMHO, we should be conservative with this release to avoid experience with few previous. There is no better marketing than good release.
+11 (eleven) Vahis -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org For additional commands, e-mail: opensuse-factory+help@opensuse.org
On Sat, Jun 20, 2009 at 08:45:56PM -0500, Rajko M. wrote:
On Saturday 20 June 2009 04:41:52 am Cristian Morales Vega wrote:
We all agree that "enough time" is needed to test changes... but nobody agrees about how much is "enough time". That's why we have a project manager... to bother him about changes we want even if there is no time and blame him about other changes that got done without time ;-)
IMHO, we should be conservative with this release to avoid experience with few previous. There is no better marketing than good release.
With the opening of Factory to the community you might find it not as conservative as you would like, with all new workflows, processes and package acceptance strategies and so on. as-needed fixes will be just a minor hickup compared to that. Ciao, Marcus -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org For additional commands, e-mail: opensuse-factory+help@opensuse.org
On 17/06/09 05:08, Stephan Kulow wrote: but surely is a lot cleaner dependency tree. That was my point ;)
And then there are tons of build failures because packages are old and/or developers don't care too much for linking and dump libraries in random order and random number into the link line and let the linker sort it out. But sometimes a library looks unneeded and is only needed later -> boom. These cases are easy to detect and disabled by changing the binutils default back for these packages.
Most of this issues, are fairly easy to fix,
* C++ code in shared libraries may depend on the correct order of initializing static objects. This hit e.g. KDE years ago, as-needed became popular enough long ago to don't make this a real problem.
So I would like to hear feedback on this. We have an alternative of course: leaving binutils as it is and only set SUSE_ASNEEDED=1 in packages where we expect real benefits. Or we set SUSE_ASNEEDED=0 in all packages creating a build problem and you guys say we test enough to make as-needed useful.
While I certainly cannot promise fix all the broken packages, please tell people to email me or this list if they have doubts about fixing this problems, this issue has to be sorted out. -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org For additional commands, e-mail: opensuse-factory+help@opensuse.org
participants (16)
-
Andreas Jaeger
-
Cristian Morales Vega
-
Cristian Rodríguez
-
David C. Rankin
-
Dominique Leuenberger
-
Greg R.
-
Joerg Mayer
-
Luis Medinas
-
Magnus Boman
-
Marcus Meissner
-
Rajko M.
-
Robert Kaiser
-
Stephan Kulow
-
Vahis
-
Vincent Untz
-
Wolfgang Rosenauer