[Mesa-dev] Possible configuration bug in the Gallium EGL state tracker

[Matt Turner]

On Wed, Oct 10, 2012 at 6:32 PM, Tzvetan Mikov <tmikov at jupiter.com> wrote:
> Hi,
> I am new to this list, so I apologize if the format or the subject is inappropriate. If so, I would appreciate any advice directing me to the appropriate place to post.
>
> I found a configuration/build time problem with the Gallium EGL state tracker. It fails in Mesa-9.0 where it used to work with Mesa-8.0.4. The problem is that -DHAVE_PTHREAD=1 is not defined anymore, but only in that specific Makefile, which causes its structures to change size and various other side effects. As far as I can tell, the problem was introduced in commit a2f28ce "build: Use AX_PTHREAD's HAVE_PTHREAD preprocessor definition".
>
> A quick hack below restoring the definition fixes the problem, though it is probably not the best way to do it, since the author of the commit probably intended something else.
>
> Thanks,
> Tzvetan
>
> diff --git a/configure.ac b/configure.ac
> index 2d922f6..b3239bd 100644
> --- a/configure.ac
> +++ b/configure.ac
> @@ -160,13 +160,13 @@ DEFINES=""
>  AC_SUBST([DEFINES])
>  case "$host_os" in
>  linux*|*-gnu*|gnu*)
> -    DEFINES="$DEFINES -D_GNU_SOURCE"
> +    DEFINES="$DEFINES -D_GNU_SOURCE -DHAVE_PTHREAD=1"
>      ;;
>  solaris*)
> -    DEFINES="$DEFINES -DSVR4"
> +    DEFINES="$DEFINES -DSVR4 -DHAVE_PTHREAD=1"
>      ;;
>  cygwin*)
> -    DEFINES="$DEFINES"
> +    DEFINES="$DEFINES -DHAVE_PTHREAD=1"
>      ;;
>  esac

Thanks for the heads-up. This is my fault. I'll fix it up and make
sure it goes into the 9.0 branch.