Instead of applying patch in git, fetch it from the local Mercurial repository

[marco-c]

Using git-cinnabar, we can replace

bugbug/scripts/commit_classifier.py

Lines 358 to 380 in e60a51b

	if self.git_repo_dir:
	patch_proc = subprocess.Popen(
	["patch", "-p1", "--no-backup-if-mismatch", "--force"],
	stdin=subprocess.PIPE,
	cwd=self.git_repo_dir,
	)
	patch_proc.communicate(patch.patch.encode("utf-8"))
	assert patch_proc.returncode == 0, "Failed to apply patch"
	subprocess.run(
	[
	"git",
	"-c",
	f"user.name={author_name}",
	"-c",
	f"user.email={author_email}",
	"commit",
	"-am",
	message,
	],
	check=True,
	cwd=self.git_repo_dir,
	)

and just fetch the commit from the local Mercurial repository where we already applied the patch.

Then, we can stop installing patch in our Docker image:

bugbug/infra/dockerfile.commit_retrieval

Line 10 in e60a51b

apt-get install -y --no-install-recommends gcc g++ curl patch && \

.

[DeepikaKaranji]

Hi @marco-c, I'd like to try and fix this issue! I hope that's okay :)

[marco-c]

Sure, feel free! Any issue is open for contribution by anyone. As a first step, read #1092.

[DeepikaKaranji]

May I know where the Mercurial repo for bugbug is?

[marco-c]

https://hg.mozilla.org/mozilla-central/

[DeepikaKaranji]

So I was able to fetch the commit from my terminal like this:

$git cinnabar fetch hg::https://hg.mozilla.org/mozilla-central/ 50244579252a229131ae04a1d2c60682611e5702

I replaced lines 358-380 with:

But I am not sure how to test this file(bugbug/scripts/commit_classifier.py), without having to run trainer.py as mentioned in the README.

[marco-c]

You will need to fetch the commit that was just applied to the local mercurial repository, not a specific commit which already exists in https://hg.mozilla.org/mozilla-central/.

[DeepikaKaranji]

I think I've not understood what a "local mercurial repo" means.
If I clone the link https://hg.mozilla.org/mozilla-central/ ,is it referred to as a local .hg repo?
How or Where do I find the 'commit that was just applied to the local mercurial repository'?
I'm really sorry if its a silly question, I'm a complete beginner :P

[marco-c]

If you look at the code of the commit classifier script mentioned in the first comment, it is cloning a Mercurial repository, then cloning a git repository using git cinnabar. Then it applies a set of patches to the locally cloned Mercurial repository and tries to apply the same set of patches to the locally cloned git repository.
Instead of trying to apply the set of patches to the locally cloned git repository, we want to fetch the patches from the locally cloned Mercurial repository, since we just applied them there.

[KhadijaKamran]

Hi @marco-c ! I am an Outreachy applicant and I want to work on this issue. Can you confirm if the set of patches are applied to Mercurial from line 352-356?

[marco-c]

@KhadijaKamran the patch is applied to the Mercurial repo at

bugbug/scripts/commit_classifier.py

Lines 376 to 380 in f990605

	hg.import_(
	patches=io.BytesIO(patch.patch.encode("utf-8")),
	message=message.encode("utf-8"),
	user=f"{author_name} <{author_email}>".encode("utf-8"),
	)

, and then to the git repo at

bugbug/scripts/commit_classifier.py

Lines 383 to 404 in f990605

	patch_proc = subprocess.Popen(
	["patch", "-p1", "--no-backup-if-mismatch", "--force"],
	stdin=subprocess.PIPE,
	cwd=self.git_repo_dir,
	)
	patch_proc.communicate(patch.patch.encode("utf-8"))
	assert patch_proc.returncode == 0, "Failed to apply patch"
	subprocess.run(
	[
	"git",
	"-c",
	f"user.name={author_name}",
	"-c",
	f"user.email={author_email}",
	"commit",
	"-am",
	message,
	],
	check=True,
	cwd=self.git_repo_dir,
	)

.
What we want to change is the way to apply the patch to the git repo (see the previous comments in this issue).

[KhadijaKamran]

Hey @marco-c! Would the following command help:
git cinnabar hg2git changeset