[FFmpeg-devel] [PATCH] build: fix make checkheaders in out-of-tree builds
gajjanag at mit.edu
Sun Jan 24 18:41:05 CET 2016
On Sun, Jan 24, 2016 at 10:14 PM, Ronald S. Bultje <rsbultje at gmail.com> wrote:
> On Sun, Jan 24, 2016 at 11:14 AM, Henrik Gramner <henrik at gramner.com> wrote:
>> On Sun, Jan 24, 2016 at 5:02 PM, Hendrik Leppkes <h.leppkes at gmail.com>
>> > Windows doesn't particularly care which slash direction you give it,
>> > so no, that changes nothing. I'm not quite sure why it fails with the
>> > include path this way, maybe some msys shenanigans.
>> > For fun and giggles, I hard-coded the correct include path in the
>> > common.mak file and that made config.h show up again of course, but
>> > the good old compiler error came back just like before.
>> > Might be msys thats interfering here and translating something it
>> > shouldn't, or double-translating something, I can't really tell just
>> > from the commands it calls.
>> > Like I said before, when initial MSVC support was build by Ronald and
>> > Martin, there were lots of troubles with absolute paths, which in the
>> > end were solved by not using them (sorry, no answers here).
>> Is there any reason why we're using absolute paths then? Can't we just
>> stick to relative paths with forward slashes since they seem to work
>> with all tools?
>> I added some code to x264's configure some time ago to convert
>> absolute paths into relative paths just in order to avoid some
>> MSYS/Cygwin headaches.
> I think Andreas' original patch removed that so in-tree and out-tree builds
> are bit-exact (i.e. the pathnames to in-tree source files are identical
> between the two build types).
> I don't know if that's worth it. Andreas seems to claim it is, others seem
> to doubt it...
FWIW, I do agree that Andreas's goals wrt the out of tree build are
worthy. To roughly the same degree you and some others dislike "ubsan
appeasement" or "theoretical changes", I don't like the GNU/Linux (or
more broadly POSIX) platform to suffer gratuitously.
Unfortunately, I lack a setup and time to peer into the make and
microsoft internals to offer some technical suggestions, beyond an
idea I proposed earlier of testing the presence of windows via some
> ffmpeg-devel mailing list
> ffmpeg-devel at ffmpeg.org
More information about the ffmpeg-devel