Re: [Pan-users] Odd issue with latest git code

[Jim Henderson]

On Sun, 13 Apr 2014 04:20:30 +0000, Duncan wrote:

>> Where does the info get pulled from for about->help and the posting
>> header?
> 
> If you search git, you'll find the commit that added that information...
> a couple years ago, actually probably three or more now. =:^)
> 
> AFAIK it uses git functionality (a git show query or some such) to set
> the value at build time.

Ah, that would explain it - the spec RPM I have doesn't specify that git 
is needed for the build, so it's probably just using what was in the 
directory when I first built it.  I should probably clear my copy out, 
get a completely clean copy.

My process may also change because OBS has now got git capabilities 
directly built into it.  I just need to figure out how to use it.

What I've been doing is:

1.  git pull
2.  tar cfvmJ pan2.tar.xz pan-0.140
3.  cp pan2.tar.xz ~/osc/myrepo (copies the xz file to my own osc build 
directory)
4.  osc build openSUSE_13.1 x86_64 (builds for this target locally)
5.  osc commit

The last command triggers a rebuild on the build service servers, so I 
end up with an RPM in my own repo on the service, allowing me to update 
it and have it integrated in my package manager.

But last week or the week before, they integrated git (or any rcs, I 
think) into the osc functionality, so in theory it should be able to pull 
directly from the repo, which saves me steps 1-3 at least (and maybe 1-4).

But I probably started with an unclean local git repo (with a build in 
the repo itself), so the variable in question was set and never got reset 
because the spec file doesn't give it git to go and update the revision 
info in the build itself.

> Which I'd guess means you're updating the repo but then checking out a
> specific commit from sometime back instead of using master-branch HEAD.
> I know nothing about Open Build Service, but certainly gentoo's various
> generation git-eclass based live-ebuilds can set a specific variable
> that will tell the ebuild to checkout and build that specific
> commit/tag/ whatever, instead of HEAD, and I'd guess you're somehow
> configured to do something similar there.
> 
> Did you ever have a build/run problem with HEAD and set it to pull from
> an earlier commit that worked... and then forget about it so it's still
> using that now rather older than intended commit?  That'd be the most
> likely explanation.
> 
> Meanwhile, if you're current, all those color patches and etc have been
> applied to master, so you'd be getting patch-reversed errors trying to
> apply them.  The fact that you're not, again indicates that you're
> building from something rather older than current master HEAD.
> 
> So I'd guess we've found the problem; now you just have to figure out
> where that stale config setting that old commit is coming from, and
> change it! =:^)

I think my thinking above will clear it up for me (at least), so I can 
look into it further tomorrow.  Looks like a spec file patch might be in 
order. :)

Jim

-- 
 Jim Henderson
 Please keep on-topic replies on the list so everyone benefits