Bug 118550 – ANGLE should be updated

• Status: RESOLVED
• Resolution: FIXED
• Component: ANGLE

[Alex Christensen]

I have a working build of WebGL on Windows that uses most of ANGLE, so we should included the additional parts of ANGLE.  Before we do this, we may as well update to the latest revision.  

Blink is currently using r2184 of ANGLE.  We're currently using r1987.  I suggest we update to the latest revision of ANGLE, r2249.

Our current differences from the ANGLE repository are that we used Bison 2.4.2 instead of Bison 2.3 to generate glslang_tab and ExpressionParser cpp and h files, we put #if defined(_MSC_VER) around #pragma warning(disable: 4718) in DependencyGraph.cpp, and we include khrplatform.h instead of KHR/khrplatform.h in ShaderLang.h.  The current revision of ANGLE used Bison 2.4.2 to generate the files, so we won't need to rerun Bison.  We may as well keep the other two changes for now.

Updating to the latest revision moves DetectRecursion.cpp and DetectRecursion.h to DetectCallDepth.cpp and DetectCallDepth.h and adds builtin_symbol_table.cpp and builtin_symbol_table.h, so I've tried to do that by hand with all build systems.  "DetectRecursion" was also in Source\WebCore\WebCore.order, and I'm not sure what that is, so I may have broken whatever that is (Stephanie - you added that).  I would not be surprised if this first patch breaks build systems, but that's what the EWS system is for.  I don't want to land a patch that breaks anything.

[Alex Christensen]

Created attachment 206429 [details]
Patch

[kov's GTK+ EWS bot]

Comment on attachment 206429 [details]
Patch

Attachment 206429 [details] did not pass gtk-ews (gtk):
Output: http://webkit-queues.appspot.com/results/931496

[Build Bot]

Comment on attachment 206429 [details]
Patch

Attachment 206429 [details] did not pass mac-wk2-ews (mac-wk2):
Output: http://webkit-queues.appspot.com/results/1026461

[Build Bot]

Comment on attachment 206429 [details]
Patch

Attachment 206429 [details] did not pass mac-ews (mac):
Output: http://webkit-queues.appspot.com/results/1058082

[Alex Christensen]

Created attachment 206566 [details]
Patch

[Build Bot]

Comment on attachment 206566 [details]
Patch

Attachment 206566 [details] did not pass mac-ews (mac):
Output: http://webkit-queues.appspot.com/results/1067030

[kov's GTK+ EWS bot]

Comment on attachment 206566 [details]
Patch

Attachment 206566 [details] did not pass gtk-ews (gtk):
Output: http://webkit-queues.appspot.com/results/1067031

[Alex Christensen]

Created attachment 206579 [details]
Patch

[kov's GTK+ EWS bot]

Comment on attachment 206579 [details]
Patch

Attachment 206579 [details] did not pass gtk-ews (gtk):
Output: http://webkit-queues.appspot.com/results/1064222

[Martin Robinson]

The build error on GTK+ is due to the combination of the bitfields with the enum. I haven't yet determined if this is a bug in gcc yet, but I did confirm that making the enums members sans bitfields fixed the issue.

[Alex Christensen]

Created attachment 206620 [details]
Patch

[EFL EWS Bot]

Comment on attachment 206620 [details]
Patch

Attachment 206620 [details] did not pass efl-ews (efl):
Output: http://webkit-queues.appspot.com/results/970772

[EFL EWS Bot]

Comment on attachment 206620 [details]
Patch

Attachment 206620 [details] did not pass efl-wk2-ews (efl-wk2):
Output: http://webkit-queues.appspot.com/results/1045681

[Early Warning System Bot]

Comment on attachment 206620 [details]
Patch

Attachment 206620 [details] did not pass qt-ews (qt):
Output: http://webkit-queues.appspot.com/results/1066395

[Early Warning System Bot]

Comment on attachment 206620 [details]
Patch

Attachment 206620 [details] did not pass qt-wk2-ews (qt-wk2):
Output: http://webkit-queues.appspot.com/results/1069137

[Alex Christensen]

Created attachment 206622 [details]
Patch

[Alex Christensen]

This is the workaround I did for getting Types.h to compile with GCC:

+#ifdef __GNUC__
+    TBasicType type;
+    TPrecision precision;
+    TQualifier qualifier;
+#else
     TBasicType type      : 6;
     TPrecision precision;
     TQualifier qualifier : 7;
+#endif

Running webgl tests on Mac doesn't add any failures with this patch, and running Safari with this patch seems to render webgl fine.

[Dean Jackson]

Comment on attachment 206622 [details]
Patch

We explicitly use the older Bison so that we don't commit code with a license that is incompatible with WebKit's terms. You'll need to regenerate the files with the older Bison.

[Dean Jackson]

i.e. use Bison 2.3, not 2.4x

Otherwise it all looks good.

[Alex Christensen]

Created attachment 206683 [details]
changes I made to ANGLE before running bison

[Alex Christensen]

Created attachment 206700 [details]
Patch

[Early Warning System Bot]

Comment on attachment 206700 [details]
Patch

Attachment 206700 [details] did not pass qt-ews (qt):
Output: http://webkit-queues.appspot.com/results/1067846

[Early Warning System Bot]

Comment on attachment 206700 [details]
Patch

Attachment 206700 [details] did not pass qt-wk2-ews (qt-wk2):
Output: http://webkit-queues.appspot.com/results/1071599

[kov's GTK+ EWS bot]

Comment on attachment 206700 [details]
Patch

Attachment 206700 [details] did not pass gtk-ews (gtk):
Output: http://webkit-queues.appspot.com/results/1073307

[Build Bot]

Comment on attachment 206700 [details]
Patch

Attachment 206700 [details] did not pass mac-wk2-ews (mac-wk2):
Output: http://webkit-queues.appspot.com/results/1071600

[Alex Christensen]

Created attachment 206705 [details]
Patch

[Alex Christensen]

Created attachment 206713 [details]
Patch

[Dean Jackson]

Comment on attachment 206713 [details]
Patch

rs=me

[WebKit Commit Bot]

Comment on attachment 206713 [details]
Patch

Clearing flags on attachment: 206713

Committed r152755: <http://trac.webkit.org/changeset/152755>

[WebKit Commit Bot]

All reviewed patches have been landed.  Closing bug.

[Dean Jackson]

Rolled out in https://trac.webkit.org/r152946

[Alex Christensen]

Created attachment 207292 [details]
Patch

[Alex Christensen]

This time I ran leaks tests and it doesn't add 4 million leaks

[Build Bot]

Comment on attachment 207292 [details]
Patch

Attachment 207292 [details] did not pass mac-ews (mac):
Output: http://webkit-queues.appspot.com/results/1092434

[EFL EWS Bot]

Comment on attachment 207292 [details]
Patch

Attachment 207292 [details] did not pass efl-ews (efl):
Output: http://webkit-queues.appspot.com/results/1082683

[Early Warning System Bot]

Comment on attachment 207292 [details]
Patch

Attachment 207292 [details] did not pass qt-ews (qt):
Output: http://webkit-queues.appspot.com/results/1172159

[Early Warning System Bot]

Comment on attachment 207292 [details]
Patch

Attachment 207292 [details] did not pass qt-wk2-ews (qt-wk2):
Output: http://webkit-queues.appspot.com/results/1177257

[EFL EWS Bot]

Comment on attachment 207292 [details]
Patch

Attachment 207292 [details] did not pass efl-wk2-ews (efl-wk2):
Output: http://webkit-queues.appspot.com/results/1092437

[Build Bot]

Comment on attachment 207292 [details]
Patch

Attachment 207292 [details] did not pass mac-wk2-ews (mac-wk2):
Output: http://webkit-queues.appspot.com/results/1082681

[Alex Christensen]

Created attachment 207296 [details]
Patch

[Early Warning System Bot]

Comment on attachment 207296 [details]
Patch

Attachment 207296 [details] did not pass qt-wk2-ews (qt-wk2):
Output: http://webkit-queues.appspot.com/results/1135460

[EFL EWS Bot]

Comment on attachment 207296 [details]
Patch

Attachment 207296 [details] did not pass efl-wk2-ews (efl-wk2):
Output: http://webkit-queues.appspot.com/results/1137488

[kov's GTK+ EWS bot]

Comment on attachment 207296 [details]
Patch

Attachment 207296 [details] did not pass gtk-ews (gtk):
Output: http://webkit-queues.appspot.com/results/1137489

[Alex Christensen]

Created attachment 207298 [details]
Patch

[kov's GTK+ EWS bot]

Comment on attachment 207298 [details]
Patch

Attachment 207298 [details] did not pass gtk-ews (gtk):
Output: http://webkit-queues.appspot.com/results/1184016

[Alex Christensen]

I'm pretty sure my patch didn't break the GTK build because the error has nothing to do with ANGLE and GTK-WK2 built successfully.  I also ran leak tests locally with this patch and it does not add 4 million leaks like the last one did.

[Alex Christensen]

Created attachment 207338 [details]
Patch

[Early Warning System Bot]

Comment on attachment 207338 [details]
Patch

Attachment 207338 [details] did not pass qt-wk2-ews (qt-wk2):
Output: http://webkit-queues.appspot.com/results/1119317

[kov's GTK+ EWS bot]

Comment on attachment 207338 [details]
Patch

Attachment 207338 [details] did not pass gtk-ews (gtk):
Output: http://webkit-queues.appspot.com/results/1183293

[Alex Christensen]

Created attachment 207343 [details]
Patch

[Dean Jackson]

Comment on attachment 207343 [details]
Patch

rs=me

[WebKit Commit Bot]

Comment on attachment 207343 [details]
Patch

Rejecting attachment 207343 [details] from commit-queue.

Failed to run "['/Volumes/Data/EWS/WebKit/Tools/Scripts/webkit-patch', '--status-host=webkit-queues.appspot.com', '--bot-id=webkit-cq-02', 'land-attachment', '--force-clean', '--non-interactive', '--parent-command=commit-queue', 207343, '--port=mac']" exit_code: 2 cwd: /Volumes/Data/EWS/WebKit

Last 500 characters of output:
/git.webkit.org/WebKit
   576a5fb..9ce069c  master     -> origin/master
Partial-rebuilding .git/svn/refs/remotes/origin/master/.rev_map.268f45cc-cd09-0410-ab3c-d52691b4dbfc ...
Currently at 153061 = 576a5fb847db0f1a9dd8fb3db2d257c8f2d817c5
r153062 = 9ce069c97e2f045384ba83f8d8c96113820779de
Done rebuilding .git/svn/refs/remotes/origin/master/.rev_map.268f45cc-cd09-0410-ab3c-d52691b4dbfc
First, rewinding head to replay your work on top of it...
Fast-forwarded master to refs/remotes/origin/master.

Full output: http://webkit-queues.appspot.com/results/1158437

[Alex Christensen]

r153064