Re: Git Test Coverage Report (May 6, 2020)

From: Taylor Blau <me@ttaylorr.com>
Date: May 6, 2020, 17:45 UTC
Message-ID: <20200506174524.GE6078@syl.local>
In-Reply-To: <4a47a215-f636-3f94-17f5-33f13140bdc7@gmail.com>

Hi Stolee,

On Wed, May 06, 2020 at 11:19:38AM -0400, Derrick Stolee wrote:

> Here is today's test coverage report.

Thanks as always for running these. I personally find them very helpful.

Show 23 quoted lines

> Thanks,
> -Stolee
>
>
> [1] https://derrickstolee.github.io/git-test-coverage/reports/2020-05-06-commits.txt
> [2] https://derrickstolee.github.io/git-test-coverage/reports/2020-05-06.txt
> [3] https://derrickstolee.github.io/git-test-coverage/reports/2020-05-06.htm
>
> ---
>
> pu	c0f7ae5a7ea5cf8246270fcf1853726167ca8ef4
> jch	fffcd002d6befa77784a44046b986934b87e6c3b
> next	55c56c9c57da646085d27e61a3420043978ab4a1
> master	07d8ea56f2ecb64b75b92264770c0a664231ce17
> master@{1}	86ab15cb154862b6fa5cc646dac27532f881e1fb
>
>
> Uncovered code in 'pu' not in 'jch'
> --------------------------------------------------------
>
> Taylor Blau	199d758e builtin/commit-graph.c: dereference tags in builtin
> builtin/commit-graph.c
> 199d758e 244) progress = start_delayed_progress(

I think that this one is fine. In the commit message, we even say something to the effect of "hey, we're probably never going to actually show this progress meter, at least in the case of a not-horrendous amount of input with '--stdin-commits'."

I'm not even sure what the standard procedures are for testing progress meters from 't'. I'm inclined to just let this one be.

Show 6 quoted lines

> Uncovered code in 'master' not in 'master@{1}'
> --------------------------------------------------------
>
> Taylor Blau	37b9dcab shallow.c: use '{commit,rollback}_shallow_file'
> builtin/receive-pack.c
> 37b9dcab 896) rollback_shallow_file(the_repository, &shallow_lock);

I'm lukewarm on whether or not this should be tested. The same commit, 37b9dcab, is testing another location of 'rollback_shallow_file'. If 'rollback_shallow_file' were defective, we'd expect to see a crash with 'git -c fetch.writeCommitGraph=true fetch ...', which we know from Elijah's original report.

So, I think that we have high-enough confidence that 'rollback_shallow_file' is working as intended.

> Taylor Blau	fdbde82f builtin/commit-graph.c: introduce split strategy 'no-merge'
> commit-graph.c
> fdbde82f 1788) break;

This is stopping us from trying to roll up commit-graph chains when they cross an alternate. It's probably worth adding a test for this (although this line has probably always been uncovered, and is only showing up again because fdbde82f indented these lines by surrounding the loop with a new if statement.)

Thanks, Taylor