If errorpage.html is unloaded before this HTML and related resources aren't
loaded completed, this issue seems to occur.
When GeckoView-junit test runs `loadExternalDenied` test after finishing
`loadHTTPSOnlyInSubframe` test, since GeckoView doesn't handle subframe error
(ex. Bug 1692578), this situation seems to occurs. So, as workaround, we
should set L10N arguments to avoid test failure.
This failure occurs on my workstation frequency, but after applying this fix,
no failure even if I run over 1 hour with --run-until-failure.
Differential Revision: https://phabricator.services.mozilla.com/D174932
Experimentation confirmed years ago that about:welcome shoudln't be treated as blank for urlbar focus, and no need for additional experimentation. Also fix a bug where about:welcome was treated as a blank/empty page resulting in about:preferences incorrectly replacing it.
Differential Revision: https://phabricator.services.mozilla.com/D174940
glxtest is run later when Firefox already spawns threads. Recently glxtest runs in forked process
which doesn't work correctly in multi-thread environment, so we need to move glxtest to different binary file
and launch it as stand alone code.
Differential Revision: https://phabricator.services.mozilla.com/D173486
- Implement fire_vaapi_process() which launch VA-API test utility on given DRM device.
- Implement GfxInfo::GetDataVAAPI() which gets VA-API test results
- Run VA-API tests when FEATURE_HARDWARE_VIDEO_DECODING is probed and only if it's enabled by GfxInfo.
Differential Revision: https://phabricator.services.mozilla.com/D171995
Right now we fire glxtest on every Firefox start, even if we going to update, restart or ping running remote instance.
When we're running on system with broken/unstable gfx drivers (drivers/glx freezes or crashes) every such action is delayed or coredumps are generated on systems.
In this patch we launch glx test proces later if we know we need it.
Depends on D168650
Differential Revision: https://phabricator.services.mozilla.com/D168651
Most of the calls to `InitFunctionEnvironmentsObjects` on Speedometer are from
Baseline JIT code. This patch optimizes that by allocating these environment
objects directly from JIT code, similar to what Ion does.
The micro-benchmark in bug 1826078 improves from 55 ms to 23 ms with `--no-ion`.
Differential Revision: https://phabricator.services.mozilla.com/D174889
`JitScript` stores template objects that are used by `WarpBuilder` for function environment
allocation. This patch makes it possible to use the same objects also for Baseline JIT
compilation in the next patch.
`CachedIonData` contains just a pointer and a small struct, and the pointer had to move
out of there, so the patch removes this indirection and stores both of these directly
in `JitScript` with `Maybe<>`.
Differential Revision: https://phabricator.services.mozilla.com/D174888
DrawTargetWebgl renders a path by uploading vertex data to the back of
a large VBO using glBufferSubData then issuing a draw call, orphaning
the buffer when it becomes full. This results in many glBufferSubData
calls being interleaved with draw calls. On Mali GPUs this causes
severe performance issues as the driver is unable to determine that
any pending draw calls do not reference the updated region of the
buffer, and therefore must create a copy of the buffer for each
update.
However, since *we* know that we never overwrite a region that is
referenced by a submitted draw call, we can force the driver to avoid
making these copies. We do so by adding a new function
UnsynchronizedBufferSubData(), which acts like BufferSubData so long
as this rule is followed. Internally, this uses glMapBufferRange with
GL_MAP_UNSYNCHRONIZED_BIT, allowing the driver to omit the extraneous
copies.
Differential Revision: https://phabricator.services.mozilla.com/D174685
Additionally check for the new content modal and standalone version for WAIT_FOR_MIGRATION_CLOSE by using observer notifications from preferences dialog close, new wizard actor, old wizard uninit.
Differential Revision: https://phabricator.services.mozilla.com/D174692
The previous code passed |translation| to PostTranslate() directly,
now we pass it through the TranslateThumb() helper which expects
it in CSS pixels (in particular not including the device scale).
Differential Revision: https://phabricator.services.mozilla.com/D174938
The |translation| variable was redeclared in the branch, shadowing the
variable in the function's top-level scope, but it's the latter that
was actually used to transform the thumb.
The patch also removes a shadowing |asyncScroll| variable, though that
one was harmless.
Differential Revision: https://phabricator.services.mozilla.com/D174936
be -> 588caf79b9ddca246b0c9e1560ec4d5321464ad8
fy-NL -> a7908ee1e8ff2a106046fa09aa763aeabd28b810
it -> 22146fa18280c6b5abec012a6771a933bd270a7c
lo -> e9fbc84d9081023b45aa4a0a66c5a8f38c3e2910
sk -> 136b72d42cb784ef721e523cfdc05a0be6ce898a
Use parsed atoms to store popovertarget attributes as more effective mechanism to search for ID match in attr associated elements implementation. Followup from D173337.
Differential Revision: https://phabricator.services.mozilla.com/D174786