Created attachment 181766 [details] Fix build graphics/dri with VAAPI option enabled
Thanks for submitting the patch. I have confirmed the BUILD_DEPEND on v4l_compat is needed for VAAPI, but that alone appears to be sufficient to build with the option enabled in poudriere. What prompted the addition of VAAPI_LDFLAGS?
*** Bug 218674 has been marked as a duplicate of this bug. ***
Eric Camachat, Please, next time follow the bug reporting convention: Component Description Individual Port or Package Prefix the issue Summary field with category/port: so that MAINTAINER is Assigned or CC'd automatically. Apart from auto assigning or auto CC'ing the maintainer - it helps others to not file duplicate issues. About the patch: with it dri got successfully built for me by synth (which it failed to do without it).
(In reply to Matthew Rezny from comment #1) Linker complained some functions not found, I just added all libs that contents those function into VAAPI_LDFLAGS.
(In reply to John Smith from comment #3) Sure, I will follow the convention.
Wouldn't it make more sense to add multimedia/v4l_compat as a RUN_DEPENDS to multimedia/libva instead? ${LOCALBASE}/include/va/va_backend.h has #include <linux/videodev2.h>, but since v4l_compat is only listed as a BUILD_DEPENDS in libva, anything including va_backend.h (like dri) in a clean environment will be broken.
(In reply to Jason E. Hale from comment #6) That is a good suggestion. I have not looked too close at libva, but that sounds like the right place to fix it if all dependents need the header.
(In reply to Matthew Rezny from comment #7) > (In reply to Jason E. Hale from comment #6) > > That is a good suggestion. I have not looked too close at libva, but that > sounds like the right place to fix it if all dependents need the header. Hi, Since I maintain libva, and noticed this discussion (by chance) I'm adding myself here. I need to look at this a little to really understand what's happening. I have no objection to adding a dependency in principle but let me study the situation and understand what's happening. I'll report back ASAP.
The suggested fix of adding a RUN_DEPENDS on libva actually fixes the issue. I'll commit the update to libva after a little further testing. Thanks for pointing this out!
A commit references this bug: Author: madpilot Date: Wed Apr 19 14:31:49 UTC 2017 New revision: 438845 URL: https://svnweb.freebsd.org/changeset/ports/438845 Log: Add a v4l_compat as RUN_DEPENDS too, as dri requires include files from this port when including optional VAAPI support. PR: 218642 Submitted by: Eric Camachat <eric@camachat.org> Suggested by: jhale@ Changes: head/multimedia/libva/Makefile
Hi, does r438845 completely fix this problem? Can I close this bug report as fixed?
(In reply to Guido Falsi from comment #11) Yes, it's fixed, go head and close it.