Request to move "gitpkgv.bbclass" from meta-oe to oe-core

The gitpkgv class is very helpful when you want packages to have a consistent incremental version number, when build machines cannot talk to eachother (ruling out using PRSERVER for example) other than through repositories.

However, this small gem is present in meta-oe and not in the core. I now have several layers that pull in all of meta-oe just for being able to number GIT versions incrementally.

Could gitpkgv be moved into oe-core? (I'll be happy to write and submit the patch.) Or, is there an alternative that does something similar?

Kind regards,
Mike.

Met vriendelijke groet / kind regards,

Mike Looijmans

TOPIC Embedded Systems
Eindhovenseweg 32-C, NL-5683 KH Best
Postbus 440, NL-5680 AK Best
Telefoon: (+31) (0) 499 33 69 79
Telefax: (+31) (0) 499 33 69 70
E-mail: [email protected]
Website: www.topic.nl

Please consider the environment before printing this e-mail

Topic zoekt gedreven (embedded) software specialisten!
http://topic.nl/vacatures/topic-zoekt-software-engineers/

In partial reply to my own question, this has been discussed 2 years ago, but apparently nothing ever came out of it.

http://lists.openembedded.org/pipermail/openembedded-core/2012-February/056050.html

So once more the request to be able to get sane version strings for git repos. Can we please just move gitpkgv. That's like five minutes work, and if anyone (e.g. me) feels like building something better, he or she is free do submit that and get rid of the gitpkgv class once it's obsolete.

Quoted 9 lines · Mike Looijmans

In partial reply to my own question, this has been discussed 2 years ago, but apparently nothing ever came out of it.

http://lists.openembedded.org/pipermail/openembedded-core/2012-February/056050.html

So once more the request to be able to get sane version strings for git repos. Can we please just move gitpkgv. That's like five minutes work, and if anyone (e.g. me) feels like building something better, he or she is free do submit that and get rid of the gitpkgv class once it's obsolete.

I do feel quite strongly that this is something which should get improved in the fetcher itself. That is why I've pushed back on the patch since if I take it into OE-Core, there is even less incentive to fix the real problem...

Quoted 11 lines · Richard Purdie

Quoted 9 lines · Mike Looijmans

In partial reply to my own question, this has been discussed 2 years ago, but apparently nothing ever came out of it.

http://lists.openembedded.org/pipermail/openembedded-core/2012-February/056050.html

So once more the request to be able to get sane version strings for git repos. Can we please just move gitpkgv. That's like five minutes work, and if anyone (e.g. me) feels like building something better, he or she is free do submit that and get rid of the gitpkgv class once it's obsolete.

Quoted 4 lines

I do feel quite strongly that this is something which should get improved in the fetcher itself. That is why I've pushed back on the patch since if I take it into OE-Core, there is even less incentive to fix the real problem...

Mike I am also an user of gitpkgv but I must to agree with Richard here. Moving this to OE-Core will just make it worse. It is way better to prepare a patch and add support for it on the fetcher code of BitBake.

Quoted 21 lines · Otavio Salvador

On Thu, Oct 9, 2014 at 4:56 PM, Richard Purdie

Quoted 13 lines · <[email protected]> wrote:

Quoted 11 lines · Mike Looijmans

In partial reply to my own question, this has been discussed 2 years ago, but apparently nothing ever came out of it.

http://lists.openembedded.org/pipermail/openembedded-core/2012-February/
056050.html

So once more the request to be able to get sane version strings for git repos. Can we please just move gitpkgv. That's like five minutes work, and if anyone (e.g. me) feels like building something better, he or she is free do submit that and get rid of the gitpkgv class once it's obsolete.

Quoted 4 lines

I do feel quite strongly that this is something which should get improved in the fetcher itself. That is why I've pushed back on the patch since if I take it into OE-Core, there is even less incentive to fix the real problem...

Quoted 4 lines

Mike I am also an user of gitpkgv but I must to agree with Richard here. Moving this to OE-Core will just make it worse. It is way better to prepare a patch and add support for it on the fetcher code of BitBake.

Give me some pointers on what you're talking about here, and I'll be happy to try and implement it. Currently I have no idea as to what is "wrong" with the gitpkgv approach, and I also have no clue as to what it is you want to have implemented in the git fetcher.

Hi Mike,

Quoted 27 lines · On Friday 10 October 2014 07:36:25 Mike Looijmans wrote:

Quoted 21 lines · Otavio Salvador

On Thu, Oct 9, 2014 at 4:56 PM, Richard Purdie

Quoted 13 lines · <[email protected]> wrote:

Quoted 11 lines · Mike Looijmans

In partial reply to my own question, this has been discussed 2 years ago, but apparently nothing ever came out of it.

http://lists.openembedded.org/pipermail/openembedded-core/2012-February/
056050.html

So once more the request to be able to get sane version strings for git repos. Can we please just move gitpkgv. That's like five minutes work, and if anyone (e.g. me) feels like building something better, he or she is free do submit that and get rid of the gitpkgv class once it's obsolete.

Quoted 4 lines

I do feel quite strongly that this is something which should get improved in the fetcher itself. That is why I've pushed back on the patch since if I take it into OE-Core, there is even less incentive to fix the real problem...

Quoted 4 lines

Mike I am also an user of gitpkgv but I must to agree with Richard here. Moving this to OE-Core will just make it worse. It is way better to prepare a patch and add support for it on the fetcher code of BitBake.

Quoted 4 lines

Give me some pointers on what you're talking about here, and I'll be happy to try and implement it. Currently I have no idea as to what is "wrong" with the gitpkgv approach, and I also have no clue as to what it is you want to have implemented in the git fetcher.

The code would go into the fetcher itself i.e. lib/bb/fetch2/git.py and lib/bb/fetch2/__init__.py.

I'm not sure whether it would be a case of setting a variable in order to have the value of SRCPV within OE (which calls bb.fetch2.get_srcrev() ) return the "git describe" style version instead of the current one, or whether we would preserve the separate GITPKGVTAG variable for that purpose. (The GITPKGV case is already handled by the current behaviour of SRCPV, right?)

FWIW there is an open bug covering this issue:
https://bugzilla.yoctoproject.org/show_bug.cgi?id=2872

Quoted 38 lines · Paul Eggleton

Hi Mike,

Quoted 27 lines · On Friday 10 October 2014 07:36:25 Mike Looijmans wrote:

Quoted 21 lines · Otavio Salvador

On Thu, Oct 9, 2014 at 4:56 PM, Richard Purdie

Quoted 13 lines · <[email protected]> wrote:

Quoted 11 lines · Mike Looijmans

In partial reply to my own question, this has been discussed 2 years ago, but apparently nothing ever came out of it.

http://lists.openembedded.org/pipermail/openembedded-core/2012-February/
056050.html

So once more the request to be able to get sane version strings for git repos. Can we please just move gitpkgv. That's like five minutes work, and if anyone (e.g. me) feels like building something better, he or she is free do submit that and get rid of the gitpkgv class once it's obsolete.

Quoted 4 lines

I do feel quite strongly that this is something which should get improved in the fetcher itself. That is why I've pushed back on the patch since if I take it into OE-Core, there is even less incentive to fix the real problem...

Quoted 4 lines

Mike I am also an user of gitpkgv but I must to agree with Richard here. Moving this to OE-Core will just make it worse. It is way better to prepare a patch and add support for it on the fetcher code of BitBake.

Quoted 4 lines

Give me some pointers on what you're talking about here, and I'll be happy to try and implement it. Currently I have no idea as to what is "wrong" with the gitpkgv approach, and I also have no clue as to what it is you want to have implemented in the git fetcher.

The code would go into the fetcher itself i.e. lib/bb/fetch2/git.py and lib/bb/fetch2/__init__.py.

Quoted 8 lines

I'm not sure whether it would be a case of setting a variable in order to have the value of SRCPV within OE (which calls bb.fetch2.get_srcrev() ) return the "git describe" style version instead of the current one, or whether we would preserve the separate GITPKGVTAG variable for that purpose. (The GITPKGV case is already handled by the current behaviour of SRCPV, right?)

FWIW there is an open bug covering this issue:
https://bugzilla.yoctoproject.org/show_bug.cgi?id=2872

I followed that link a few clicks deeper and noticed that Richard Purdie suggested adding another interface method, but I couldn't quite understand what he meant here.

I figured that "sortable_revision" was the method that I needed to delve into. However, it turned out that this one is called about 50 times when building a single package, and i cannot really run any git commands at that spot it seems, the current directory is not inside the repo.

So that's a fail (my trial code from git.py with some extra logger output below).

def sortable_revision(self, ud, d, name):
"""
Return a sortable revision number by counting commits in the history """
rev = self._build_revision(ud, d, name) logger.warn("git.sortable_revision url=%s" % (ud.url)) localpath = ud.localpath
rev_file = os.path.join(localpath, "oe-gitpkgv_" + rev)
if not os.path.exists(localpath):
logger.warn("git.sortable_revision non-existent: %s" % (localpath)) commits = None
else:
if not os.path.exists(rev_file) or not os.path.getsize(rev_file):
from pipes import quote
commits = bb.fetch2.runfetchcmd(
"git rev-list %s -- 2> /dev/null | wc -l" %
(quote(rev)),

                         d, quiet=True).strip().lstrip('0')
                 if commits:
                     os.unlink(rev_file)
                     open(rev_file, "w").write("%d\n" % int(commits))
             else:
                 commits = open(rev_file, "r").readline(128).strip()
         logger.warn("git.sortable_revision commits=%s\n" % (commits))
         if commits:
             return False, "%s+%s" % (commits, rev[:7])
         else:
             return True, str(rev)