metadata_scm.bbclass: get the metadata revision for current HEAD
PATCH 1/1 metadata_scm.bbclass: get the metadata revision for current HEAD
Currently we report wrong revision when we build from branches not based on master because we only fetch the HEAD value which is remote master This patch makes it to report the HEAD of whatever branch one has checked out.
--- classes/metadata_scm.bbclass | 3 ++- 1 files changed, 2 insertions(+), 1 deletions(-)
classes/metadata_scm.bbclass +2 -1
@@ -70,7 +70,8 @@ def base_get_metadata_git_branch(path, d): return "<unknown>" def base_get_metadata_git_revision(path, d): - rev = os.popen("cd %s; PATH=%s git show-ref HEAD 2>/dev/null" % (path, d.getVar("PATH", 1))).read().split(" ")[0].rstrip() + branch = os.path.basename(base_get_metadata_git_branch(path, d)) + rev = os.popen("cd %s; PATH=%s git show-ref HEAD %s 2>/dev/null" % (path, d.getVar("PATH", 1), branch)).read().split(" ")[0].rstrip() if len(rev) != 0: return rev return "<unknown>"
Signature · 1 line
1.7.1
Quoted 5 lines · Khem Raj
Currently we report wrong revision when we build from branches not based on master because we only fetch the HEAD value which is remote master This patch makes it to report the HEAD of whatever branch one has checked out.
This doesn't make sense. The HEAD ref always points to the current checked out commit or branch, it is *not* specific to master.
Signature · 5 lines
Christopher Larson clarson at kergoth dot com Founder - BitBake, OpenEmbedded, OpenZaurus Maintainer - Tslib Senior Software Engineer, Mentor Graphics
Martin Jansa
Acked-byQuoted 10 lines · Chris Larson
Quoted 5 lines · Khem Raj
Currently we report wrong revision when we build from branches not based on master because we only fetch the HEAD value which is remote master This patch makes it to report the HEAD of whatever branch one has checked out.
This doesn't make sense. The HEAD ref always points to the current checked out commit or branch, it is *not* specific to master.
No it does make sense, because it fixes my <unknown> METADATA_REVISION :).
in my R/W checkout I have it right even before patch
bitbake@jama ~/oe $ git show-ref HEAD b3246d96069fd11caee42ec6ebcbf6dca2d62449 refs/remotes/origin/HEAD
OE om-gta02@shr ~/oe $ git remote -v
jama /OE/dev/ (fetch)
jama /OE/dev/ (push)
....
origin [email protected]:openembedded (fetch) origin [email protected]:openembedded (push)
but in my "work" checkout, where I have this R/W used as origin bitbake@jama ~/dev $ git show-ref HEAD -> nothing
OE om-gta02@shr ~/dev $ git remote -v
gitorious [email protected]:~jama/angstrom/jama-shr-experimental.git (fetch) gitorious [email protected]:~jama/angstrom/jama-shr-experimental.git (push)
....
origin /OE/oe/ (fetch)
origin /OE/oe/ (push)
Regards,
Signature · 1 line
Martin 'JaMa' Jansa jabber: [email protected]
Quoted 14 lines · Martin Jansa
Quoted 3 lines · Chris Larson
Quoted 1 line · Khem Raj
Currently we report wrong revision when we build from branches not
based
Quoted 6 lines
Quoted 4 lines
on master because we only fetch the HEAD value which is remote master This patch makes it to report the HEAD of whatever branch one has checked out.
This doesn't make sense. The HEAD ref always points to the current
checked
Quoted 1 line
out commit or branch, it is *not* specific to master.
Quoted 8 lines
No it does make sense, because it fixes my <unknown>
METADATA_REVISION:).Acked-by Martin Jansa <[email protected]>in my R/W checkout I have it right even before patch
bitbake@jama ~/oe $ git show-ref HEAD
b3246d96069fd11caee42ec6ebcbf6dca2d62449refs/remotes/origin/HEAD
No, this is broken. It shouldn't be determining the METADATA_REVISION from a ref you don't even have checked out. Right now it's pulling the HEAD ref from upstream, it should be pulling the local HEAD ref. So NACK on this patch, the metadata revision code needs to be fixed to either not use show-ref or pass the necessary args to not operate based on a HEAD of a remote.
Signature · 5 lines
Christopher Larson clarson at kergoth dot com Founder - BitBake, OpenEmbedded, OpenZaurus Maintainer - Tslib Senior Software Engineer, Mentor Graphics
Quoted 26 lines · Chris Larson
Quoted 14 lines · Martin Jansa
Quoted 3 lines · Chris Larson
Quoted 1 line · Khem Raj
Currently we report wrong revision when we build from branches not
based
Quoted 6 lines
Quoted 4 lines
on master because we only fetch the HEAD value which is remote master This patch makes it to report the HEAD of whatever branch one has checked out.
This doesn't make sense. The HEAD ref always points to the current
checked
Quoted 1 line
out commit or branch, it is *not* specific to master.
Quoted 8 lines
No it does make sense, because it fixes my <unknown>
METADATA_REVISION:).Acked-by Martin Jansa <[email protected]>in my R/W checkout I have it right even before patch
bitbake@jama ~/oe $ git show-ref HEAD
b3246d96069fd11caee42ec6ebcbf6dca2d62449refs/remotes/origin/HEAD
Quoted 6 lines
No, this is broken. It shouldn't be determining the
METADATA_REVISIONfrom a ref you don't even have checked out. Right now it's pulling the HEAD ref from upstream, it should be pulling the local HEAD ref. So NACK on this patch, the metadata revision code needs to be fixed to either not use show-ref or pass the necessary args to not operate based on a HEAD of a remote.
whats wrong with this patch. or what is it that we need additionally.
Quoted 11 lines
--
Christopher Larson
clarson at kergoth dot com
Founder - BitBake, OpenEmbedded, OpenZaurus
Maintainer - Tslib
Senior Software Engineer, Mentor Graphics
_______________________________________________
Openembedded-devel mailing list
[email protected]
http://lists.linuxtogo.org/cgi-bin/mailman/listinfo/openembedded-devel
Quoted 6 lines · Khem Raj
Quoted 5 lines
Quoted 2 lines
bitbake@jama ~/oe $ git show-ref HEAD
b3246d96069fd11caee42ec6ebcbf6dca2d62449refs/remotes/origin/HEADNo, this is broken. It shouldn't be determining the
METADATA_REVISIONfrom
Quoted 8 lines
Quoted 1 line
a ref you don't even have checked out. Right now it's pulling the HEAD
ref
Quoted 4 lines
from upstream, it should be pulling the local HEAD ref. So NACK on this patch, the metadata revision code needs to be fixed to either not use show-ref or pass the necessary args to not operate based on a HEAD of a remote.
whats wrong with this patch. or what is it that we need additionally.
The issue is, show-ref shows all refs which match the supplied search string. show-ref HEAD will show all the HEAD refs, both local and the ones in all your remotes, which isn't what we want for METADATA_REVISION. I'm testing a patch which changes it to use rev-parse to get the commit hash for HEAD rather than show-ref. Thanks for catching this, by the way, I don't want to sound unappreciative, its just that this patch is more of a workaround, rather than a fix for the root problem. :)
Signature · 5 lines
Christopher Larson clarson at kergoth dot com Founder - BitBake, OpenEmbedded, OpenZaurus Maintainer - Tslib Senior Software Engineer, Mentor Graphics
Quoted 15 lines · Chris Larson
Quoted 13 lines · Khem Raj
Quoted 5 lines
Quoted 2 lines
bitbake@jama ~/oe $ git show-ref HEAD
b3246d96069fd11caee42ec6ebcbf6dca2d62449refs/remotes/origin/HEADNo, this is broken. It shouldn't be determining the
METADATA_REVISIONfrom
Quoted 1 line
a ref you don't even have checked out. Right now it's pulling the HEAD
ref
Quoted 4 lines
from upstream, it should be pulling the local HEAD ref. So NACK on this patch, the metadata revision code needs to be fixed to either not use show-ref or pass the necessary args to not operate based on a HEAD of a remote.
Quoted 8 lines
Quoted 1 line
whats wrong with this patch. or what is it that we need additionally.
The issue is, show-ref shows all refs which match the supplied search string. show-ref HEAD will show all the HEAD refs, both local and the ones in all your remotes, which isn't what we want for
METADATA_REVISION. I'm testing a patch which changes it to use rev-parse to get the commit hash for HEAD rather than show-ref.
FWIW you could also use git show-ref --head <branch>
Thanks for catching this, by the way, I don't
Quoted 13 lines
want to sound unappreciative, its just that this patch is more of a workaround, rather than a fix for the root problem. :) --
Christopher Larson
clarson at kergoth dot com
Founder - BitBake, OpenEmbedded, OpenZaurus
Maintainer - Tslib
Senior Software Engineer, Mentor Graphics
_______________________________________________
Openembedded-devel mailing list [email protected]
http://lists.linuxtogo.org/cgi-bin/mailman/listinfo/openembedded-devel
Quoted 9 lines · Khem Raj
Quoted 2 lines
The issue is, show-ref shows all refs which match the supplied search string. show-ref HEAD will show all the HEAD refs, both local and the
ones
Quoted 2 lines
in all your remotes, which isn't what we want for
METADATA_REVISION. I'm testing a patch which changes it to use rev-parse to get the commit hashfor
Quoted 1 line
HEAD rather than show-ref.
FWIW you could also use git show-ref --head <branch>
Yeah, that will work if the <branch> is the full path including refs/heads/, but it's not using the right plumbing tool for the job. show-ref is designed to *list* refs in the repo, not resolve a single one.
Signature · 5 lines
Christopher Larson clarson at kergoth dot com Founder - BitBake, OpenEmbedded, OpenZaurus Maintainer - Tslib Senior Software Engineer, Mentor Graphics