| View previous topic :: View next topic |
| Author |
Message |
l0l
n00b
Joined: 10 Jul 2008
Posts: 43
|
 Posted: Tue Jun 05, 2012 10:53 am Post subject: firefox & pgo & nvidiactl, once again |
 |
|
There is previous discussion here: https://forums.gentoo.org/viewtopic-t-912182-highlight-nvidiactl.html which is marked as solved,
but the src_compile() is still broken:
| Code: |
src_compile() {
232 if use pgo; then
233 addpredict /root
234 addpredict /etc/gconf
235 # Reset and cleanup environment variables used by GNOME/XDG
236 gnome2_environment_reset
237
238 # Firefox tries to use dri stuff when it's run, see bug 380283
239 shopt -s nullglob
240 cards=$(echo -n /dev/dri/card* | sed 's/ /:/g')
241 if test -n "${cards}"; then
242 # FOSS drivers are fine
243 addpredict "${cards}"
244 else
245 cards=$(echo -n /dev/ati/card* /dev/nvidiactl* | sed 's/ /:/g')
246 if test -n "${cards}"; then
247 # Binary drivers seem to cause access violations anyway, so
248 # let's use indirect rendering so that the device files aren't
249 # touched at all. See bug 394715.
250 export LIBGL_ALWAYS_INDIRECT=1
251 fi
252 fi
253 shopt -u nullglob
254
255 CC="$(tc-getCC)" CXX="$(tc-getCXX)" LD="$(tc-getLD)" \
256 MOZ_MAKE_FLAGS="${MAKEOPTS}" \
257 Xemake -f client.mk profiledbuild || die "Xemake failed"
258 else
259 CC="$(tc-getCC)" CXX="$(tc-getCXX)" LD="$(tc-getLD)" \
260 MOZ_MAKE_FLAGS="${MAKEOPTS}" \
261 emake -f client.mk || die "emake failed"
262 fi
263
264 }
|
because this:
| Code: |
241 if test -n "${cards}"; then
|
does not work when:
| Code: |
ls /dev/dri/card*
ls: cannot access /dev/dri/card*: No such file or directory
|
line 241 can be changed to something like this to check if the file exists:
| Code: |
241 if [ $(ls /dev/dri/card* 2> /dev/null | wc -l) != "0" ]; then
|
and the same applies to line 246. |
|
| Back to top |
|
 |
Hu
Moderator
Joined: 06 Mar 2007
Posts: 21624
|
| Posted: Wed Jun 06, 2012 1:40 am Post subject: |
|
|
| The command you gave seems to be a roundabout method of achieving the same result. When nullglob is enabled and there are no DRI cards, that echo will produce no output, which is the desired result. Your command will also produce a result of zero lines. Could you explain why you think the existing code does not work correctly? |
|
| Back to top |
|
|
l0l
n00b
Joined: 10 Jul 2008
Posts: 43
|
| Posted: Wed Jun 06, 2012 8:53 am Post subject: |
|
|
| Hu wrote: | | The command you gave seems to be a roundabout method of achieving the same result. When nullglob is enabled and there are no DRI cards, that echo will produce no output, which is the desired result. Your command will also produce a result of zero lines. Could you explain why you think the existing code does not work correctly? |
Hi, yes you are absolutely right, both commands produce the same result. Sorry for that, it's my fault because I did not set nullglob when testing, completely forgot about that.
So the code works correctly, the problem is somwhere else then, because when I compiled firefox 13 there was the same access violation error in /dev/nvidiactl, so I compiled firefox 13
the second time with FEATURES="-sandbox" which worked fine.
After investigating I found that the problem is that 'export LIBGL_ALWAYS_INDIRECT=1' does not work for me, I still get direct rendering for some reason. |
|
| Back to top |
|
|
|
|
You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot vote in polls in this forum
|
|
Portage & Programming
