weird commits on top of develop

85 views
Skip to first unread message

Vincent Delecroix

unread,
Feb 26, 2023, 5:13:48 AM2/26/23
to sage-devel
Hello,

Why do we have commits on top of develop after the release commits? Namely

commit 52a81cbd161ef4d5895325657c88a68590ea1d3b
Author: Dima Pasechnik <di...@pasechnik.info>
Date: Fri Feb 24 21:15:42 2023 +0000

Revert "add missing # optional - gap3"

This reverts commit c017a6a3d68f0aca8cb63ceba52fe451a64881a3.

commit c017a6a3d68f0aca8cb63ceba52fe451a64881a3
Author: Dima Pasechnik <di...@pasechnik.info>
Date: Fri Feb 24 11:57:32 2023 +0000

add missing # optional - gap3

to let CI pass

Best
Vincent

Dima Pasechnik

unread,
Feb 26, 2023, 5:16:42 AM2/26/23
to sage-devel
it was a mistake on my side - I thought develop branch was broken, but it was not.

--
You received this message because you are subscribed to the Google Groups "sage-devel" group.
To unsubscribe from this group and stop receiving emails from it, send an email to sage-devel+...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/sage-devel/CAGEwAAkF50jNxVth6io%2BCErWu9mWFW4L2D%2BERqQVcNf7aPjp1A%40mail.gmail.com.

Vincent Delecroix

unread,
Feb 26, 2023, 5:19:24 AM2/26/23
to sage-...@googlegroups.com
I do not understand why develop isn't protected against anybody but
the release manager pushes.
> To view this discussion on the web visit https://groups.google.com/d/msgid/sage-devel/CAAWYfq0XTpPcxuKuE11ChKpx0jcAjsxGRVsrD%2BLDeQO-X4RxnQ%40mail.gmail.com.

Dima Pasechnik

unread,
Feb 26, 2023, 10:46:37 AM2/26/23
to sage-devel
sometimes Volker misses things which make CI break, and such a release is pretty much useless, in the sense that you can't gauge anything from the CI results on PRs made on top of it. Or an important CI or other infrastructure improvement doesn't make it into the release.

In previous release cycles we few times pushed necessary changes in develop to fix such things.
We do have OK from Volker for these fixes.

This time I got confused and pushed a totally irrelevant change (which did not break develop, it just had no effect on anything at all).

Dima 




Vincent Delecroix

unread,
Feb 26, 2023, 11:29:29 AM2/26/23
to sage-...@googlegroups.com
In other words, everytime you got confused we will have a develop
branch not aligned with the most recent tag and a bunch of extra
commits cancelling each other?
> To view this discussion on the web visit https://groups.google.com/d/msgid/sage-devel/CAAWYfq0U9GvqhM6qy8katYmoXE-xeGW3z41moGcicSucVzHwpw%40mail.gmail.com.

Volker Braun

unread,
Feb 26, 2023, 3:36:42 PM2/26/23
to sage-devel
I agree that we should turn on branch protection, it is rather easy to accidentally push to develop.

Its also not ok to push what you think is a CI fix without it going to review, we had some irregularities to support the github transition but that is over now. 



Dima Pasechnik

unread,
Feb 26, 2023, 3:42:21 PM2/26/23
to sage-...@googlegroups.com
On Sun, Feb 26, 2023 at 4:29 PM Vincent Delecroix
<20100.d...@gmail.com> wrote:
>
> In other words, everytime you got confused we will have a develop
> branch not aligned with the most recent tag and a bunch of extra
> commits cancelling each other?

I don't think I really can constructively reply to an email pitched in
such an offensive way.
But, whatever, sorry for offending your sense of the beautiful, if
this is what you want to hear.
> To view this discussion on the web visit https://groups.google.com/d/msgid/sage-devel/CAGEwAAk-ork4G3BXoC%3DkVOQSSQ%3D0Hb86sxeMw7cB2vbN%2BJ3x6g%40mail.gmail.com.

Matthias Koeppe

unread,
Feb 26, 2023, 4:54:29 PM2/26/23
to sage-devel
Branch protection is on, but currently admins can bypass it. This can be tightened in the config here if so desired:

Tobias Diez

unread,
Feb 27, 2023, 9:32:08 AM2/27/23
to sage-devel
The branch protection rules don't allow to make an exception for a single person (the release manager). What can be configured though is that all pushes to develop (from everybody, including admins) have to go through a PR. This would mean that the merge and release commits also has to go through a PR. Maybe this is a good idea anyway, since then the github ci checks that everything is okay (relative to the setup tests) and would prevent the need for these "fix the ci" commits.

John Cremona

unread,
Feb 27, 2023, 10:01:37 AM2/27/23
to sage-...@googlegroups.com
We could also have a branch called release which is separate from develop.  Normal developers would make PRs to be merged into develop like now, but the release manager could make PRs from develop to release.  That's the way we do things in the LMFDB (with different names for the branches).  So developers could wait until there was a new commit to release before updating unless they were happy to live (a little more) dangerously and pulling from develop itself.

John

Reply all
Reply to author
Forward
0 new messages