[gdal-dev] Clang-format

[Mateusz Loskot]

On Sat, 10 Oct 2020 at 23:39, Even Rouault <even.rouault at spatialys.com> wrote:
>
> > However, I found it strange that there is no automatic formatting with
> > clang-format for C++ code.
> >
> > In this page
> > https://gdal.org/development/rfc/rfc69_cplusplus_formatting.html it is
> > suggesting to use it. That would be great! Enabling it in your favourite
> > editor makes the code nicer, without any extra work. And more consistent!
>
> > About "when" to apply it (reformatting *all* the existing code), I think a
> > good moment could be just before creating a "big" release branch, like 3.2.
> > Then the backport of bug fixes would be easy (otherwise it is a nightmare).
> >
> > What do you think?
>
>
> Was discussed at that time in
> https://lists.osgeo.org/pipermail/gdal-dev/2017-May/thread.html#46560
>
> I would be quite in favor of the principle
>
> A few points I've in mind:
>
> - I'm slightly worried about clang-format output changing between versions of the tool (based on my experience with PROJ, I can tell you that they are different between 3.8, which is the one we used for PROJ, and 10), which would be an issue as we cannot be certain that our set of contributors to have different versions. Not sure if having a custom .clang-format file (possibly a 'standard' one that we put in our tree to make output stable) would help ? This would have to be investigated & tested.

I think, it would require us to stick to a particular clang-format version
that is allowed to (re)format GDAL.
Contributors who want to use clang-format would have to use
the decided version only.

I suspect, higher versions of clang-format systematically suffer
from less and less behaviour discrepancies, so it may be sensible
to pick one of the later versions, instead of 3.8.

Best regards,
-- 
Mateusz Loskot, http://mateusz.loskot.net