[PATCH libevdev] build: set aux-dir

Peter Hutterer peter.hutterer at who-t.net
Mon Nov 4 22:49:21 CET 2013


On Mon, Nov 04, 2013 at 04:12:17PM -0500, Gaetan Nadon wrote:
> On 13-11-04 01:55 PM, David Herrmann wrote:
> > I'm on rolling-release, so 2.69 it is. I know how subdir-objects work,
> > the thing is, we include our source files in ./libevdev/Makefile.am
> > and ./test/Makefile.am and during "make distclean" ./libevdev/.deps/
> > got removed before distclean is called in ./test/, thus failing "make
> > distcheck" as it cannot find its .Po dependencies..
> >
> > If you know how to solve that, I'd be happy to get hints. But
> > otherwise I'm just too lazy to fix Peter's recursive Makefiles ;) If
> > we'd switch to one global top_srcdir makefile, these problems would go
> > away by itself. 
> I don't think so. Automake works the same way even if you cram all the
> source code in one makefile.
> > But maybe we can just live without setting
> > "subdir-objects". I don't really care that much.
> I have not seen subdir-objects in libevdev. Xorg has hundreds of
> packages the size of libevdev and there seems to be no need for it.

libevdev uses the source twice, once for the library and then again for the
tests. I did this instead of using the .la file because when a test fails it
is a matter of minutes to run it in gdb and figure out the exact cause. with
the .la file I have to run gdb through libtool, which doesn't like
re-compilations, etc. if you can figure out a way to statically compile the
sources in through a .a file and avoid the whole libtool mess I'd be quite
grateful.

Cheers,
   Peter

> The tests cannot run successfully on my box, so I disabled the building
> of the tests. The target 'distcheck' runs ok and a tarball gets created.
> I'll try to weed out the failing test cases to try to reproduce your
> scenario. Can you post the make distcheck failure?


More information about the Input-tools mailing list