1963568 - `mach try --push-to-lando` with a git-cinnabar based repo doesn't use the right `hg` base commit (can't push anything to Try)

Status: RESOLVED FIXED

(Firefox Build System :: General, defect)

Description

[William Durand [:willdurand]]

I am using a git-cinnabar based repo and I cannot push to Try anymore. This started today before the Git migration. I am seeing this:

$ ./mach try --preset webextensions --push-to-lando --artifact
estimates: Runs 107 tasks (81 selected, 26 dependencies)
estimates: Total task duration 8:30:03
estimates: In the shortest 35% of durations
estimates: Should take about 1:33:05 (Finished around 2025-04-30 15:14)
Note: `--push-to-lando` is now the default behaviour of `mach try`.
Note: Use `--push-to-vcs` to push changes to try directly.
Auth0 token validated.
Using fddef8c3f434b0bd4129568ed2e74d32e2a190ce as the hg base commit.
[...]

fddef8c3f434b0bd4129568ed2e74d32e2a190ce is a git commit, and not a mercurial commit.

When I revert the patch for Bug 1963450, I am getting this:

./mach try --preset webextensions --push-to-lando --artifact
estimates: Runs 107 tasks (81 selected, 26 dependencies)
estimates: Total task duration 8:30:03
estimates: In the shortest 35% of durations
estimates: Should take about 1:33:05 (Finished around 2025-04-30 15:17)
Note: `--push-to-lando` is now the default behaviour of `mach try`.
Note: Use `--push-to-vcs` to push changes to try directly.
Auth0 token validated.
Using 272b2cc681a93c88cfe833458f46750219793f3d as the hg base commit.
[...]

Also:

$ git cinnabar git2hg fddef8c3f434b0bd4129568ed2e74d32e2a190ce
272b2cc681a93c88cfe833458f46750219793f3d

So it looks like Bug 1963450 regressed mach try with git-cinnabar. Not sure whether this is expected or not, and whether the Git migration will solve this or not, but not being able to push anything to Try is not cool.

Comment 1

[BugBot [:suhaib / :marco/ :calixte]]

:sfink, since you are the author of the regressor, bug 1963450, could you take a look? Also, could you set the severity field?

For more information, please visit BugBot documentation.

Comment 2

[Connor Sheehan [:sheehan]]

We backed out the change from bug 1963450 to resolve this for now.

Comment 3

[Steve Fink [:sfink] [:s:]]

Gah! My apologies, I didn't think I was changing the git path at all. Though re-reading the patch, I see that I was switching git-cinnabar to post a (cinnabar) git commit instead of an hg rev. And then I only tested a firefox git repo push to try. :-(

I will of course hold off on changing anything for now. But what is the desired end state here? Should a git-cinnabar repo be sending the base as hg or git? I thought from reading somewhere else that either gecko-dev or firefox git commit hashes were going to be used and the server would translate gecko-dev -> firefox, but I must have gotten that wrong. (Partly I'm asking because I may take a stab at creating a test for this.)

Comment 4

[Mike Hommey [:glandium]]

The end state is that a git repo should send a git commit hash, probably, but not it if's plain git-cinnabar or gecko-dev, only the new repo. The code for artifact builds has a function that returns which kind of repo a git repo is, that could reused here.

Comment 5

[BugBot [:suhaib / :marco/ :calixte]]

Set release status flags based on info from the regressing bug 1963450

Comment 6

[Steve Fink [:sfink] [:s:]]

Fixed by the backout.