public inbox for openembedded-core@lists.openembedded.org
 help / color / mirror / Atom feed
From: Randolph Sapp <rs@ti.com>
To: Paul Barker <paul@pbarker.dev>, <rs@ti.com>, <raj.khem@gmail.com>,
	<richard.purdie@linuxfoundation.org>,
	<mathieu.dubois-briand@bootlin.com>, <alex@linutronix.de>,
	<otavio@ossystems.com.br>, <kexin.hao@windriver.com>
Cc: <afd@ti.com>, <detheridge@ti.com>, <denis@denix.org>,
	<reatmon@ti.com>, <openembedded-core@lists.openembedded.org>,
	<vijayp@ti.com>
Subject: Re: [oe-core][PATCHv2] reproducible: fix git SOURCE_DATE_EPOCH randomness
Date: Thu, 19 Feb 2026 19:26:13 -0600	[thread overview]
Message-ID: <DGJEC61AKA7H.8KGAM5Z75OBQ@ti.com> (raw)
In-Reply-To: <e745402979f61ca6710e44b2116c6485ca0484ff.camel@pbarker.dev>

On Thu Feb 19, 2026 at 9:01 AM CST, Paul Barker wrote:
> On Tue, 2026-02-17 at 14:01 -0600, Randolph Sapp via
> lists.openembedded.org wrote:
>> From: Randolph Sapp <rs@ti.com>
>> 
>> Anything that defines multiple git sources should have the largest value
>> taken when calculating the SOURCE_DATE_EPOCH for a package.
>> 
>> The previous iteration actually introduced some degree of randomness, as
>> it would stop on the first git repository reported by os.walk, which
>> does not assure any specific ordering by default.
>> 
>> Signed-off-by: Randolph Sapp <rs@ti.com>
>> ---
>> 
>> v2: Use os.walk method as opposed to glob to avoid infinite recursion when
>> navigating symbolic links
>> 
>>  meta/lib/oe/reproducible.py | 63 ++++++++++++++++---------------------
>>  1 file changed, 27 insertions(+), 36 deletions(-)
>> 
>> diff --git a/meta/lib/oe/reproducible.py b/meta/lib/oe/reproducible.py
>> index 0270024a83..c58db48fb1 100644
>> --- a/meta/lib/oe/reproducible.py
>> +++ b/meta/lib/oe/reproducible.py
>> @@ -74,52 +74,43 @@ def get_source_date_epoch_from_known_files(d, sourcedir):
>>          bb.debug(1, "SOURCE_DATE_EPOCH taken from: %s" % newest_file)
>>      return source_date_epoch
>> 
>> -def find_git_folder(d, sourcedir):
>> -    # First guess: UNPACKDIR/BB_GIT_DEFAULT_DESTSUFFIX
>> -    # This is the default git fetcher unpack path
>> +def find_git_folders(d, sourcedir):
>>      unpackdir = d.getVar('UNPACKDIR')
>> -    default_destsuffix = d.getVar('BB_GIT_DEFAULT_DESTSUFFIX')
>> -    gitpath = os.path.join(unpackdir, default_destsuffix, ".git")
>> -    if os.path.isdir(gitpath):
>> -        return gitpath
>> -
>> -    # Second guess: ${S}
>> -    gitpath = os.path.join(sourcedir, ".git")
>> -    if os.path.isdir(gitpath):
>> -        return gitpath
>> -
>> -    # Perhaps there was a subpath or destsuffix specified.
>> -    # Go looking in the UNPACKDIR
>> -    for root, dirs, files in os.walk(unpackdir, topdown=True):
>> -        if '.git' in dirs:
>> -            return os.path.join(root, ".git")
>> +    git_folders = []
>> 
>> -    for root, dirs, files in os.walk(sourcedir, topdown=True):
>> -        if '.git' in dirs:
>> -            return os.path.join(root, ".git")
>> +    for mainpath in (sourcedir, unpackdir):
>> +        for root, dirs, _ in os.walk(mainpath, topdown=True):
>> +            if ".git" in dirs:
>
> Do we need to add handling for git submodules? In submodules, '.git' is
> a file instead of a directory.

Is there any way for a submodule to have a newer commit date than the parent
repository?

>> +                git_folders.append(os.path.join(root, ".git"))
>
> We should change this to `git_folders.append(root)` (see below).
>
>> 
>> -    bb.warn("Failed to find a git repository in UNPACKDIR: %s" % unpackdir)
>> -    return None
>> +    if not git_folders:
>> +        bb.warn("Failed to find any git repository in UNPACKDIR or S")
>> +
>> +    return git_folders
>> 
>>  def get_source_date_epoch_from_git(d, sourcedir):
>>      if not "git://" in d.getVar('SRC_URI') and not "gitsm://" in d.getVar('SRC_URI'):
>>          return None
>> 
>> -    gitpath = find_git_folder(d, sourcedir)
>> -    if not gitpath:
>> -        return None
>> +    # Get an epoch from all valid git repositoies
>> +    sources_dates = []
>> +    for gitpath in find_git_folders(d, sourcedir):
>> +        # Check that the repository has a valid HEAD; it may not if subdir is used
>> +        # in SRC_URI
>> +        p = subprocess.run(['git', '--git-dir', gitpath, 'rev-parse', 'HEAD'], stdout=subprocess.PIPE, stderr=subprocess.STDOUT)
>
> Using '--git-dir' does not set the path to the worktree correctly. This
> may work, but it's fragile. While we're modifying things here, can we
> change find_git_folders() to return the paths of the repository roots
> instead of the .git directories? Then we can use 'git -C path ...' here,
> which is much less likely to have issues in the future?

Fair enough, seems like good future-proofing.

>> +        if p.returncode != 0:
>> +            bb.debug(1, "%s does not have a valid HEAD: %s" % (gitpath, p.stdout.decode('utf-8')))
>> +            continue
>> 
>> -    # Check that the repository has a valid HEAD; it may not if subdir is used
>> -    # in SRC_URI
>> -    p = subprocess.run(['git', '--git-dir', gitpath, 'rev-parse', 'HEAD'], stdout=subprocess.PIPE, stderr=subprocess.STDOUT)
>> -    if p.returncode != 0:
>> -        bb.debug(1, "%s does not have a valid HEAD: %s" % (gitpath, p.stdout.decode('utf-8')))
>> -        return None
>> +        bb.debug(1, "git repository: %s" % gitpath)
>> +        p = subprocess.run(['git', '-c', 'log.showSignature=false', '--git-dir', gitpath, 'log', '-1', '--pretty=%ct'],
>> +                           check=True, stdout=subprocess.PIPE)
>> +        sources_dates.append(int(p.stdout.decode('utf-8')))
>> +
>> +    if sources_dates:
>> +        return sorted(sources_dates, reverse=True)[0]
>
> Can we use `max(sources_dates)` here?
>
> Best regards,

Yeah, my bad.

- Randolph


      reply	other threads:[~2026-02-20  1:26 UTC|newest]

Thread overview: 3+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2026-02-17 20:01 [oe-core][PATCHv2] reproducible: fix git SOURCE_DATE_EPOCH randomness rs
2026-02-19 15:01 ` Paul Barker
2026-02-20  1:26   ` Randolph Sapp [this message]

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=DGJEC61AKA7H.8KGAM5Z75OBQ@ti.com \
    --to=rs@ti.com \
    --cc=afd@ti.com \
    --cc=alex@linutronix.de \
    --cc=denis@denix.org \
    --cc=detheridge@ti.com \
    --cc=kexin.hao@windriver.com \
    --cc=mathieu.dubois-briand@bootlin.com \
    --cc=openembedded-core@lists.openembedded.org \
    --cc=otavio@ossystems.com.br \
    --cc=paul@pbarker.dev \
    --cc=raj.khem@gmail.com \
    --cc=reatmon@ti.com \
    --cc=richard.purdie@linuxfoundation.org \
    --cc=vijayp@ti.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox