threads / discuss / 66645

[Outreachy] Question about check-ref-format test coverage

Subject: [Outreachy] Question about check-ref-format test coverage

## tl;dr

5 messages between Oct 23, 2026 and Oct 24, 2026.

replies: 4people: 3as markdown or json

Fien Dora· Oct 23, 2026, 15:29 UTC · lore

Hello Everyone, I'm called Fien Dora and outreachy applicant for the December 2026 Cohort. I'm interested in the git project "Improve how command arguments and options are scanned and parsed." I'm new to open source generally but i'm ready to learn and contribute

I've been getting familiar with the git codebase, I've successfully cloned and built Git from the source on my Mac, ran the parse-options(96 test) and check-ref-format tests (107 tests), all passed.

While I was exploring builtin/check-ref-format.c, I noticed that Git supports --allow-onelevel and --no-allow-onelevel, but I couldn't find any tests of --no-allow-onelevel in the t/ directory.

i tried running
./git check-ref-format --allow-onelevel foo
./git check-ref-format --allow-onelevel --no-allow-onelevel foo
it returned with an exit code of 0 and 1 respectively.

I've also gone through the list archives and saw an earlier outreachy work converting check-ref-format to use parse-options.

Would adding tests for --allow-onelevel, particularly how it works with --no-allow-onelevel be a good contribution for my Outreachy microproject?

I look forward for any guidance
Harald Nordgren· Oct 23, 2026, 15:53 UTC · re: Fien Dora · lore

Re: [Outreachy] Question about check-ref-format test coverage

Show 8 quoted lines
> i tried running
>
> ./git check-ref-format --allow-onelevel foo
>
> ./git check-ref-format --allow-onelevel --no-allow-onelevel foo
>
>
> it returned with an exit code of 0 and 1 respectively.

`--allow-onelevel´ and ´--no-allow-onelevel´ are contradictory options, and shouldn't be run together, so that's likely why it failed with exit code 1.

This is true for any `--xyz´ and ´--no-xyz´ options.
Harald
Phillip Wood· Oct 24, 2026, 09:52 UTC · re: Harald Nordgren · lore

Re: [Outreachy] Question about check-ref-format test coverage

On 23/10/2026 16:53, Harald Nordgren wrote:
Show 11 quoted lines
>>
>> ./git check-ref-format --allow-onelevel foo
>>
>> ./git check-ref-format --allow-onelevel --no-allow-onelevel foo
>>
>> it returned with an exit code of 0 and 1 respectively.
> > `--allow-onelevel´ and ´--no-allow-onelevel´ are contradictory
> options, and shouldn't be run together, so that's likely why it failed
> with exit code 1.
>
> This is true for any `--xyz´ and ´--no-xyz´ options.
Our commands allow "--no-xyz" to override "--xyz" which is useful when an alias includes "--xyz" and you want to use the alias without that option. The second command above exits 1 because "foo" is a one-level ref and we've passed "--no-allow-onelevel" which overrides the earlier "--allow-onelevel"
Thanks
Phillip
Fien Dora· Oct 24, 2026, 18:37 UTC · re: Phillip Wood · lore

Re: [Outreachy] Question about check-ref-format test coverage

Hi Phillip,
> It sounds like a useful project to me, it is small and self contained
> and improves our codebase.
Show 6 quoted lines
>> Our commands allow "--no-xyz" to override "--xyz" which is useful
>> when
>> an alias includes "--xyz" and you want to use the alias without that
>> option. The second command above exits 1 because "foo" is a one-level
>> ref and we've passed "--no-allow-onelevel" which overrides the earlier
>> "--allow-onelevel"

Thank you for the feedback and for clarifying how the options override each other.

I'll work on adding the tests and share the patch for review.
Phillip Wood· Oct 24, 2026, 09:48 UTC · re: Fien Dora · lore

Re: [Outreachy] Question about check-ref-format test coverage

Hi Fien
On 23/10/2026 16:29, Fien Dora wrote:
Show 9 quoted lines
> Hello Everyone,
> I'm called Fien Dora and outreachy applicant for the December 2026
> Cohort. I'm interested in the git project "Improve how command
> arguments and options are scanned and parsed."
> I'm new to open source generally but i'm ready to learn and contribute
> > I've been getting familiar with the git codebase, I've successfully
> cloned and built Git from the source on my Mac, ran the
> parse-options(96 test) and check-ref-format tests (107 tests), all
> passed.
That's good
> While I was exploring builtin/check-ref-format.c, I noticed that Git
> supports --allow-onelevel and --no-allow-onelevel, but I couldn't find
> any tests of --no-allow-onelevel in the t/ directory.
Yes, "git grep -e --no-allow-onelevel" shows we don't have any tests for that, adding one would be useful.
> i tried running
> > ./git check-ref-format --allow-onelevel foo
> > ./git check-ref-format --allow-onelevel --no-allow-onelevel foo
> > it returned with an exit code of 0 and 1 respectively.
That's expected because --no-allow-onelevel comes later on the command line it overrides --allow-onelevel and so we exit 1 because the ref does not contain a '/' character.
> I've also gone through the list archives and saw an earlier outreachy
> work converting check-ref-format to use parse-options.
It's great that you've checked the archives, be sure to bear in mind the feedback that was given when you're writing your patch.
> Would adding tests for --allow-onelevel, particularly how it works
> with --no-allow-onelevel be a good contribution for my Outreachy
> microproject?
It sounds like a useful project to me, it is small and self contained and improves our codebase.
Thanks
Phillip

← back to recent threads