[Mesa-dev] [PATCH 2/2] intel/genxml: automake: drop Makefile dependency for the _xml.h target

Matt Turner mattst88 at gmail.com
Fri Oct 28 21:21:23 UTC 2016

On Fri, Oct 28, 2016 at 1:42 PM, Emil Velikov <emil.l.velikov at gmail.com> wrote:
> On 28 October 2016 at 20:55, Emil Velikov <emil.l.velikov at gmail.com> wrote:
>> On 28 October 2016 at 19:18, Lionel Landwerlin
>> <lionel.g.landwerlin at intel.com> wrote:
>>> On 28/10/16 19:12, Emil Velikov wrote:
>>>> From: Emil Velikov <emil.velikov at collabora.com>
>>>> There is nothing in here that would require it.
>>>> Cc: mesa-stable at lists.freedesktop.org
>>>> Cc: Lionel Landwerlin <lionel.g.landwerlin at intel.com>
>>>> Signed-off-by: Emil Velikov <emil.velikov at collabora.com>
>>>> ---
>>>> Not sure why we got that dependency in the first place, Lionel ?
>>> Some of the Makefile's content ends up in the header (the last echo line of
>>> this patch).
>>> If someone changes that, it's nice not to have to cleanup all the intel
>>> directory
>> Valid point. At the same time, this applies to every rule [we have in
>> mesa] and we don't bother anywhere else.
> Upon some testing/thinking I think we should nuke it.
> Reason being that as one runs ./configure the Makefile is generated.
> Thus the files will be generated regardless if a) we're a release
> tarball where the files are already present and b) we do (or don't)
> have the tool to regenerate them to begin with.

No one is doing development from tarballs, so that seems completely
irrelevant to me. If you're thinking about the case where a distro
would patch the build system... that's up to them to make work.

FWIW, I think Lionel raises a good point. Since some of the contents
of the generated file come from the Makefile itself, the Makefile
should be a dependency.

To address your point that we don't list that dependency elsewhere for
generated files -- most places don't need it since they just execute a
python script. Those cases *do* however list the python script.

I only see a couple of other trivial looking cases (in the anv and
radv vulkan Makefiles) that echo something into the resulting file. In
both cases, it's just a timestamp and nothing we'd ever want to
change, so it seems fine to not list the Makefile as a dependency.

More information about the mesa-dev mailing list