Re: Git Test Coverage Report (October 11)

[Derrick Stolee <stolee@gmail.com>]

On 10/24/2019 11:25 AM, Torsten Bögershausen wrote:
> On Wed, Oct 23, 2019 at 02:07:20PM -0400, Derrick Stolee wrote:
>> On 10/23/2019 1:00 PM, Torsten Bögershausen wrote:
>>> On Fri, Oct 11, 2019 at 09:33:11AM -0400, Derrick Stolee wrote:
>>>> Here is today's test coverage report. The usual report format is
>>>> available online [1], [2]. The report listed below is a new format
>>>> that groups lines by the commit that introduced them [3]. Thanks
>>>> Peff for the feedback on that idea.
>>>>
>>>
>>> []
>>>>
>>>> Torsten Bögershausen	ebb8d2c9 mingw: support UNC in git clone file://server/share/repo
>>>> connect.c
>>>> ebb8d2c9 921) path = host - 2; /* include the leading "//" */
>>>>
>>>
>>> I actually looked into this one, and my understanding is that the code path
>>> makes only sense for windows and is only tested on Windows in t5500.
>>> (Linux/Unix/POSIX don't use UNC path names starting with "//" )
>>>
>>> How can we avoid those "not covered by test" warnings?
>>>
>>> One solution could be to use
>>>
>>> #ifndef has_dos_drive_prefix
>>> #define has_dos_drive_prefix(a) 0
>>> #endif
>>>
>>> in git-compat-util.h and hope that the compiler is smart enough
>>> to optimize away that line of code.
>>>
>>> Another way could be to have #ifdefs in connect.c, so that it
>>> is clear "this is Windows only".
>>>
>>> Or make a comment for the "cover report" saying "not covered".
>>>
>>> Are there any good or better thoughts on this ?
>>
>> One way to avoid this is to add ignored lines to the test-coverage
>> repo [1]. These only work if the exact contents match on a specific
>> line number, but can be a way to stop noise in the short-term.
>>
>> For example, I added a few lines to ignore in commit-graph.c [2],
>> but I haven't added ignored lines in a while.
>>
>> I'm happy to take a PR including the lines you want to ignore, or
>> I could take inventory of the lines in the current report before regenerating
>> a test for -rc1.
>>
>> Thanks,
>> -Stolee
>>
>> [1] https://github.com/derrickstolee/git-test-coverage
>>
>> [2] https://github.com/derrickstolee/git-test-coverage/blob/master/ignored/commit-graph.c
> 
> I added a PR as suggested.
> One thing, that came into my mind:
> 
> Would it make sense to loosen the condition:
> 921:path = host - 2; /* include the leading "//" */
> 
> Remove the line number:
> host - 2; /* include the leading "//" */
> 
> That would assume, that the line is unique within the file,
> (can be checked with unique) .
> It can give a more robust handling
> when lines are added in the file and file numbers change,
> but the content is the same.

I'll consider making the line number optional. The reason I put
the numbers there was so we could have

	994:return 1;

and that would check a particular error-check path, but not
ALL places with "return 1;".

Your line is particularly unique, so ignoring all lines with
that text should be fine.

Thanks!
-Stolee