Re: [PATCH v2] meson: Tolerate errors from git ls-files --deduplicate

["Martin Storsjö" <martin@martin.st>]

[-- Attachment #1: Type: text/plain, Size: 2094 bytes --]

On Fri, 1 Aug 2025, Patrick Steinhardt wrote:

> On Fri, Aug 01, 2025 at 12:42:52PM +0300, Martin Storsjö wrote:
>> On Fri, 1 Aug 2025, Patrick Steinhardt wrote:
>>
>>> On Fri, Aug 01, 2025 at 10:56:22AM +0300, Martin Storsjö wrote:
>>>> index 9bc1826cb6..9b519e6eed 100644
>>>> --- a/meson.build
>>>> +++ b/meson.build
>>>> @@ -694,9 +694,12 @@ third_party_excludes = [
>>>>
>>>>  headers_to_check = []
>>>>  if git.found() and fs.exists(meson.project_source_root() / '.git')
>>>> -  foreach header : run_command(git, '-C', meson.project_source_root(), 'ls-files', '--deduplicate', '*.h', third_party_excludes, check: true).stdout().split()
>>>> -    headers_to_check += header
>>>> -  endforeach
>>>> +  ls_headers = run_command(git, '-C', meson.project_source_root(), 'ls-files', '--deduplicate', '*.h', third_party_excludes, check: false)
>>>> +  if ls_headers.returncode() == 0
>>>> +    foreach header : ls_headers.stdout().split()
>>>> +      headers_to_check += header
>>>> +    endforeach
>>>> +  endif
>>>>  endif
>>>
>>> Yup, this looks reasonable to me. We could have an `else` branch that
>>> warns about the command failing, for example like this:
>>>
>>>    warning("could not find headers: " + ls_headers.stderr())
>>
>> This would work - however the output from ls_headers.stderr() is fairly long
>> (if you try running e.g. "git ls-files --foobar", you'll get a 37 line
>> listing of potential options); it's rather distracting for what's otherwise
>> a fairly minor build configuration issue.
>>
>> Using ls_headers.stderr().split('\n')[0] works and just gets us this:
>>
>>     ../meson.build:703: WARNING: could not find headers: error: unknown option `deduplicate'
>>
>> However I wonder if it's worth it, or if it just makes the meson file
>> potentially more brittle? (E.g. what if split() returns an array of 0
>> elements? Not sure if that's possible though...)
>
> True. Maybe we can just not include stderr at all but say:
>
>    could not list headers, disabling static analysis targets
>
> That's probably sufficient.

Thanks, that sounds reasonable!

// Martin