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For years I've been told that it is obsolete.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