[cairo] Build system b0rkage and pthread

Behdad Esfahbod behdad at behdad.org
Wed Nov 17 10:28:28 PST 2010

On 08/11/10 17:35, Chris Wilson wrote:
> On Wed, 11 Aug 2010 17:01:38 -0400, Behdad Esfahbod <behdad at behdad.org> wrote:
>> > Hi,
>> > 
>> > Kinda in a rant mode.
>> > 
>> > Two specific issues:
>> > 
>> >   - This commit of Chris: 7f114b781f5c530d57530e5f76402e41cdabac6b
>> > Is very wrong for many reasons, least of which because it modifies
>> > Makefile.am.gtk-doc which we copy from gtk-doc and should not be modified.
> Makefile.am.gtk-doc doesn't say that, how are we supposed to know?
>> > I'm not sure why the commit was needed.  The commit message doesn't make sense
>> > to me.
> make distcheck fails because it (at least the version of autoconf that I
> was using...) does chmod a-w on _build/ and the original
> invocation of gtk-doc tries to modify the files within _build/, fails
> and distcheck complains loudly.

I finally got to take a deeper look into this.

You really meant autoconf does chmod a-w on _dist/, and not _build/, right?
I have a hard time believing that gtk-doc is broken in that way.  Afterall,
tens or hundreds of projects are using it and pass distcheck just fine.

The second part, however, the one about moving cairo-supported-features.h from
srcdir to builddir doesn't make sense to me.  cairo-supported-features.h is a
static file.  The CAIRO_CONFIG_COMMANDS() macro doesn't try to recreate the
file if the contents didn't change.  So, if you try to distcheck and it fails
because of a failure to change cairo-supported-features.h, it means that your
dist is broken and has a stale cairo-supported-features.h.  That shouldn't
happen.  That was by design.


More information about the cairo mailing list