Talk:Gentoo git workflow

From Gentoo Wiki
Jump to:navigation Jump to:search
This is a Talk page - please see the documentation about using talk pages. Add newer comments below older ones, sign comments using four tildes (~~~~), and indent successive comments with colons (:). Add new sections at the bottom of the page, under a heading (== ==). Please remember to mark sections as "open for discussion" using {{talk|open}}, so they will show up in the list of open discussions.

Referencing bug reports

Talk status
This discussion is still ongoing.

Various suggestions have been made on gentoo-dev on how to include bug report refernces in commit messages. For now, there is no consensus on the format. Should all be listed here? Chithanh (talk) 01:25, 12 August 2015 (UTC)

I have summed up the results and they look like

reference the bug only in the summary: 1
don't make any of this mandatory: 1
"Gentoo-Bug: 123" or similar short form: 9
"Gentoo-Bug: <url>" or similar long form: 2-3

That, to me, suggests that we should do the following:

  • make "Gentoo-Bug: 123" mandatory (for multiple bugs, separate with ,, if that exceeds the max line length, start a new "Gentoo-Bug:" line for the rest of the items)
  • leave it to the committer if he wants to add the bug number in the summary
  • leave it to the committer if he wants to add "Gentoo-Bug-url: <url>" to the commit message description

Hasufell (talk) 11:16, 12 August 2015 (UTC)

Sping (talk): My understanding was that "(bug #123456)" was the way we have always been referencing Gentoo bugs. Fot upstream bugs, "(upstream bug #12345)" or "(gnome bug #12345)" makes sense and was used as well, I believe. That would avaid the unecessary length of "Gentoo-Bug:" with Gentoo being the by far more common case, to my impression.

Chithanh (talk) 13:09, 24 August 2015 (UTC) : As still no consensus is reached I propose to add the following to the article below Gentoo git workflow#commit message format which should cover most suggestions:

Referencing bug reports

There is no policy or consensus on how to best reference bug reports in commit messages. The following are examples of what has been suggested and/or is in use currently:

If you are referencing external bug trackers instead of Gentoo Bugzilla, it is a good idea to give the full URL.

Chithanh (talk) 13:09, 24 August 2015 (UTC)


Talk status
This discussion is still ongoing.

repoman currently does not add Signed-Off-By. Do we want this? When using git to commit, it was suggested to use both -S and -s. Chithanh (talk) 01:25, 12 August 2015 (UTC)

IMO, it is fairly useless to us. It is used by the kernel, because they toss patches around wildly through different committers/authors and have to know where that thing originates and who was working on it.

The Signed-off-by: tag indicates that the signer was involved in the
development of the patch, or that he/she was in the patch's delivery path.

We mainly try to keep committer == author (e.g. we don't cherry-pick user patches, but merge them with an enforced merge commit. And even if someone else commits, there's usually just one author), so it isn't very useful to us, because we don't have such an indirect and complicated workflow.

However, Acked-by:, Reported-by:, Tested-by:, Reviewed-by:, Suggested-by: sound useful to me.

Hasufell (talk) 11:00, 12 August 2015 (UTC)

Repoman adds Signed-off-by: when DCO_SIGNED_OFF_BY is set (see man make.conf). Or does it not work any more with git?

In 2013 there was a discussion on the nfp list about making DCOs a mandatory part of our Copyright and Attribution Policy. I'm unsure what the outcome of that discussion was, but DCO_SIGNED_OFF_BY was added to portage a couple months later, which suggests we were leaning in that direction. Maybe get in touch with the trustees to see what the status of this is.

--RyanHill (talk) 06:23, 13 August 2015 (UTC)

DCO_SIGNED_OFF_BY in make.conf still works with git and is useful for me, now that commits to user repositories must be signed.

AstroFloyd (talk) 10:50, 19 November 2018 (UTC)

PORTAGE_GPG_KEY vs. user.signingkey exclamation mark

Talk status
This discussion is done.

When using a dedicated signing subkey, PORTAGE_GPG_KEY is suggested to contain a ! at the end. Under which circumstances should this apply to user.signingkey too? Chithanh (talk) 01:25, 12 August 2015 (UTC)

I was informed by K_F on IRC that the same rules apply for both user.signingkey and PORTAGE_GPG_KEY regarding whether to put a ! or not. Chithanh (talk) 13:11, 24 August 2015 (UTC)

git config

Talk status
This discussion is done.

We instruct the developers to push signed. Why not add that to the local repository's config?

user $git config --local push.gpgSign 1

--Nicolas Bock (talk) 13:16, 19 April 2016 (UTC)