Signed-off-by: Robert Eckelmann <longnoserob@postmarketos.org>

Clayton Craft authored 4 months ago and Newbyte committed 4 months ago

This doesn't depend on an auth error to detect if a user has approved a
MR, and fixes #24.

For some reason mr.approval_state().get was 404'ing, causing #24.
[ci:skip-build]: already built successfully in CI

Previously, the string representation for ActionType would be something
like ActionType.OK instead of [OK ] as intended. This resulted in
commands like `$ mrhlpr status` getting an output like this:

    ActionType.OK Changes allowed
    ActionType.OK Clean worktree
    ActionType.NOK Rebase on master
    ActionType.NOK MR-ID in commit msgs
    ActionType.NOK Commit subjects follow format
       668526 doesn't match any regex
    ActionType.NOK Commits are signed

instead of the intended

    [OK ] Changes allowed
    [OK ] Clean worktree
    [NOK] Rebase on master
    [NOK] MR-ID in commit msgs
    [NOK] Commit subjects follow format
       668526 doesn't match any regex
    [NOK] Commits are signed

Fixes 8647dd3b

mypyc is not yet completely stable, and might fail for things that are
correct Python (see
!61 (comment 456440)),
so don't make it a requirement that it passes for merging

Doesn't require to keep in sync all architectures supported by
Alpine, like ppc64le or loongarch64.

magdesign authored 5 months ago and Pablo Correa Gomez committed 5 months ago

Fixes #26

This seems to help with more reliably getting automerge going on the new
GitLab instance … maybe.

We pass this check.

With the parent of this commit, we pass this check.

We already pass this, so might as well enable it.

This makes the values easier to tweak in the future.

And remove equivalent switches from mypy.sh script to avoid
duplication.

Robert Eckelmann authored 7 months ago and Newbyte committed 5 months ago

Co-developed-by: Luca Weiss <luca@lucaweiss.eu>

Robert Eckelmann authored 5 months ago and Luca Weiss committed 5 months ago

In order to make Sphinx be able to list the supported commands, refactor
the argument parsing code to return ArgumentParser object, and align
the function naming closer to how it works in pmbootstrap.

Co-developed-by: Luca Weiss <luca@lucaweiss.eu>

Add some spacing and indent lists correctly.

Pablo Correa Gomez authored 5 months ago and Pablo Correa Gomez committed 5 months ago

GNOME's SSH endpoints areof the form
git@ssh.gitlab.gnome.org:GNOME/Incubator/papers.git and without this
patch, the host could not be found

This also effectively reverts 7d068e20,
because now we actually use the version for something in the program. It
also avoids the problems reported in
#15 to the
best of my knowledge — I tried reproducing the bug as written in there
and did not get any build failures.

The primary purpose of this version command is for making
experimentation with mypyc easier, but I also can imagine it might be
nice for users to be able to get their version of mrtest easily.

None of the code paths actually make use of the case where it returns
None for anything other than displaying the same error message, if they
display one at all. Some just attempted to use None as if it was an
integer. As such, just display the error in mr.checked_out() to simplify
and deduplicate code.

Mostly for fun. Not sure if there's any real practical benefit. It might
be useful for pmbootstrap later, but adopting it to be compatible with
mypyc would likely be more work, so I'm using this as a playground to
learn how it works.

mypyc doesn't understand the ellipsis in this context:

    mrhlpr/mr.py:177: error: Unsupported statement in class body

And they've functionally the same anyway, so just work around it like
this.

This way we're even stricter with missing types. Let's activate it so we
don't regress.

This should never happen, so we can simplify the function signature by
typing it like this.

frontend.main() only ever returns None, so there's no point in
calling sys.exit with its return value as parameter.

This is useful to test how apk actually will behave when upgrading.

It makes no sense to allow None here given that the first usage of
origin is accessing a property of it with the dot operator, an operation
which would be guaranteed to fail on a None value.

Signed-off-by: Robert Eckelmann <longnoserob@gmail.com>

Pablo Correa Gomez authored 6 months ago and Robert Eckelmann committed 6 months ago

Fixes #21

With more modern packaging methods, the executable flag gets stripped
from the files in data. Adjust mr.py to run the scripts with python3
explicitly, doing it like that is actually better than having these
scripts that are not meant to be executed directly with the executable
flag on the filesystem.

Fixes: issue 20

Pablo Correa Gomez authored 10 months ago and Pablo Correa Gomez committed 10 months ago

Fixes #15