We're experiencing a high volume of content crashes on linux, most likely due to
free type thread unsafety seen in bug 1477444 and 1479498. This commit disables
tiling to stop the crashes for now while we investigate further.
--HG--
extra : rebase_source : 9becf3d4c5e41eb6a30ed12008ed342663660291
extra : intermediate-source : b9b178105f1bd1d794e53c2fb128bd9e505f8f32
extra : source : d28438c3e7596af6d6e482f3b227c7cb43dbbf26
To be able to ignore tests for individual capabilities we need to
parametrize these tests. geckodriver now supports setWindowRect,
but fails the proxy capability test because it is for some reason
not propagated back.
For similar reasons as for platformName, tests for configuring the
timeouts object do not belong in the same parent test as those for
response body structure.
The setWindowRect capability is mandated by the WebDriver standard
and is an indication whether the driver supports manipulating the
window dimensions and position.
This will always be true for Firefox and always false for Fennec.
Change the switch-statement to only do assertions, and replace the
v variable with any parsed values before writing it to the matched
set of capabilities in one location.
Automatic update from web-platform-testsMerge pull request #12141 from fippo/remember-firefox
webrtc: make transceiver tests work in Firefox
--
wpt-commits: 00d1189bc19907a981ee8ee9b90229121bf3ac51
wpt-pr: 12141
Automatic update from web-platform-testsRevert "html: Split semantics/tabular-data/processing-model-1/span-limits.html into two."
This reverts commit bb69db9e4daaee1b42d5026d3d4596614e3908ca.
Reason for revert: Found that both of tests still timed out.
Original change's description:
> html: Split semantics/tabular-data/processing-model-1/span-limits.html into two.
>
> The third test and the forth test are very slow. Split out the latter to
> another file.
>
> Bug: https://github.com/web-platform-tests/wpt/issues/11187
> Change-Id: I3af14e712b6b06603140666f08005fb18d6a0541
> Reviewed-on: https://chromium-review.googlesource.com/1148163
> Reviewed-by: Takayoshi Kochi <kochi@chromium.org>
> Commit-Queue: Kent Tamura <tkent@chromium.org>
> Cr-Commit-Position: refs/heads/master@{#578991}
TBR=kochi@chromium.org,tkent@chromium.org
Change-Id: I1733918daad4fcc6ed83e015397b9830443b22aa
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Bug: https://github.com/web-platform-tests/wpt/issues/11187
Reviewed-on: https://chromium-review.googlesource.com/1154569
Reviewed-by: Kent Tamura <tkent@chromium.org>
Commit-Queue: Kent Tamura <tkent@chromium.org>
Cr-Commit-Position: refs/heads/master@{#579011}
--
wpt-commits: d892efb842377ae0b73b39a0d0a80f0b9a998c8f
wpt-pr: 12226
Automatic update from web-platform-tests[WIP] Implemented a prototype H2 Server that can establish HTTP/2.0 connections on port 9000, and serve directories.
* Added event handling for ConnectionTerminated
* Added backwards compatibility with HTTP/1.1 Python handler files.
* Made it so a command line flag is needed to start the H2 server, and it is no longer started by default.
* Refactored compatibility checking when launching HTTP/2.0 server
* Created H1 and H2 handler variants, keeping shared logic in a base class
* Added a `raw_headers` field to the H2Headers object, so that the non-converted keys are exposed
* Added basic multithreading support. Requests now get placed in a thread to finish handling, whilst server listens
for more requests/windows updates.
* Created a ConnectionGuard object for the H2 connection object for thread safety as per jgraham's suggestion
--
Vendored in h2 and its dependencies
--
wpt-commits: 8ccd0cdbb51f89e1c45f16b7843dca1ffca37c91, 0e78a45d9f8ebe9d0f0978d6c2fe59a512e9865b
wpt-pr: 11806
MozReview-Commit-ID: 8qojXnrrAN0
This saves about 37 KiB of memory across the UA style sheets.
MozReview-Commit-ID: EoZnlmyWwxX
--HG--
extra : rebase_source : cd8ef0ba838618f9a4583b7d9896caa3a0602199
A couple of changes to ensure that the mochitest harness doesn't try to
start executing changes before the mochitest extension is loaded:
1. Fix the marionette driver to wait for an installed extension to
be started before returning from Addon:install
2. Wait for extension API onStartup() handlers to finish before
considering a webextension started.
MozReview-Commit-ID: 8YEdNn6s5qh
--HG--
extra : rebase_source : 67e9abadcda82d55ac73c33367ec65cdbf7b823d
Support for the _xpcom_categories property was removed in Bug 568691,
but we left a bunch of consumers behind, and it's been cargo-culted a bit more.
We should remove the remaining remnants.
Differential Revision: https://phabricator.services.mozilla.com/D2429
--HG--
extra : moz-landing-system : lando
When a transceiver is stopped, its mid should not be reused by a new transceiver.
Differential Revision: https://phabricator.services.mozilla.com/D2518
--HG--
extra : moz-landing-system : lando
Checking if file upload works for files from different folders
is outside the scope of the wdspec tests. As such those tests,
and the `createFile` fixture can be removed.
MozReview-Commit-ID: DJ2seBjinxZ
--HG--
extra : rebase_source : 89bb61fa853cafece776534603aff62257de74f6
This is another quick-and-dirty port. Some of the test logic could
probably be moved to unprivileged extension code but there's no urgent
reason to do that right now.
MozReview-Commit-ID: Af0XQbakLsx
--HG--
rename : testing/mochitest/bootstrap.js => testing/mochitest/api.js
extra : rebase_source : 3870a63d96c0b2d73ccc468393c75e17b884383f
extra : source : 604167986098726b86ea0f9252a63abab36d35c4
Summary:
The memory usage of wpt-update has always been high, and accidentially
regressed in Bug 1476053 to really problematic levels when doing a
full metadata update for all test results. This patch implements a
number of changes that reduce the peak memory usage to ~1Gb and the
heap after reading all results data to ~600Mb.
THere are several changes in this patch:
* Change from a dict {test: [(subtest, prop, run_info, result)]} to a
nested dictionary {test: {subtest: [(prop, run_info,
result)]}}. This fixes the silliness that caused the previous
regression.
* Convert all unicode data to bytestrings and then intern() the
bytestring. This allows reusing string allocations for repeated
strings, of which there are many.
* Process the test manifests upfront and then allow the manifest data
to be gc'd before starting to process the results files, so that we
are not holding on to data that is no longer required.
* Stop storing status-type results as (status, default_expected); just
look up the default_expected for a specific test if we have to
update the expectations for that test.
* Add __slots__ to all frequently-allocated classes to avoid the
overhead of constructing an __dict__ for each instance, which
typically has size overhead (see
e.g. http://book.pythontips.com/en/latest/__slots__magic.html )
* Move to using a custom compact representation for the list of
per-subtest results i.e. [(prop, run_info, results)]. This uses an
array.array of 2-byte ints to store the entries and effectively
interns the values, bit packing the representations of property to
be updated and result (if it's a test status) into 4 bits each in
the first byte, and the run info into the second byte. Properties
with a complex data representation (e.g. LSAN failures) are stored
directly, with a 0 value of the status bits indicating that the
complex value should be read.
Reviewers: ato
Tags: #secure-revision
Bug #: 1479290
Differential Revision: https://phabricator.services.mozilla.com/D2496
MozReview-Commit-ID: 7zGpdzdaqRj
Automatic update from web-platform-testsFind LayoutText containing the first letter before attaching pseudo.
We assert that we find the same text node before creating the element as
we find when attaching the first letter LayoutText. We used to attach
the ::first-letter pseudo element before finding the LayoutText from
which we get the first letter text.
We did crash in a clusterfuzz test because FirstLetterTextLayoutObject()
was confused by a combination of a grid, button, anonymous
wrappers and continuations. Instead of trying to fix all bugs in
FirstLetterTextLayoutObject(), find the LayoutText before attaching the
::first-letter to make sure we are consistent instead of chasing
clusterfuzz issues.
Reported the incorrectness for button, grid, and ::first-letter in
868380.
Bug: 859285, 868380
Change-Id: I335a32b466ab31858fb05ea5f650cf12ab674040
Reviewed-on: https://chromium-review.googlesource.com/1152982
Reviewed-by: Koji Ishii <kojii@chromium.org>
Commit-Queue: Rune Lillesveen <futhark@chromium.org>
Cr-Commit-Position: refs/heads/master@{#578993}
--
wpt-commits: 3442f9d9d984af97c015dabbf1e59dde9585cca4
wpt-pr: 12215
Automatic update from web-platform-testshtml: Split semantics/tabular-data/processing-model-1/span-limits.html into two.
The third test and the forth test are very slow. Split out the latter to
another file.
Bug: https://github.com/web-platform-tests/wpt/issues/11187
Change-Id: I3af14e712b6b06603140666f08005fb18d6a0541
Reviewed-on: https://chromium-review.googlesource.com/1148163
Reviewed-by: Takayoshi Kochi <kochi@chromium.org>
Commit-Queue: Kent Tamura <tkent@chromium.org>
Cr-Commit-Position: refs/heads/master@{#578991}
--
wpt-commits: 1e0465e0b37af31c2f261c3cfb3291afa143c1af
wpt-pr: 12155
Automatic update from web-platform-testsRemove Shared Worker WPT for WebUSB
This change removes the WebUSB on Shared Worker web platform tests,
since the feature is not yet implemented in any browser.
Bug: 841510
Change-Id: I6cd09c245edcd69799b39701b82ba1510caf845e
Reviewed-on: https://chromium-review.googlesource.com/1151752
Reviewed-by: Chong Zhang <chongz@chromium.org>
Reviewed-by: Reilly Grant <reillyg@chromium.org>
Commit-Queue: Ovidio Henriquez <odejesush@chromium.org>
Cr-Commit-Position: refs/heads/master@{#578479}
--
wpt-commits: ba4921d054808a44c2966507c19861ca729a76a2
wpt-pr: 12200
Automatic update from web-platform-tests[OOR-CORS] Add a wpt for CORS with sandboxed-iframe
CORSURLLoader uses network::ResourceRequest::request_initiator which
ignores iframe sandboxing, which is a bug. Let's add a simple test for
the behavior.
Bug: 867834
Change-Id: Ia45113503f98b3d18e31c785e703d29eee658d08
Reviewed-on: https://chromium-review.googlesource.com/1151167
Commit-Queue: Yutaka Hirano <yhirano@chromium.org>
Reviewed-by: Kinuko Yasuda <kinuko@chromium.org>
Cr-Commit-Position: refs/heads/master@{#578636}
--
wpt-commits: b1f0b037511dbfb6ce9801c57fecc7efed944482
wpt-pr: 12192
Automatic update from web-platform-testsUpdate Chrome WebDriver capabilities to use vendor prefix (#11649)
The Chrome WebDriver implementation uses `goog:` for Chrome-specific
settings in its capabilites. The capabilites used in the current test
suite does not use this prefix. This commit resolves that issue.
--
wpt-commits: 312f233727faaeb58269f4cb2ea67a534e35402b
wpt-pr: 11649
Kris Maglione
Ryan Hunt
Andreas Tolfsen
Boris Zbarsky
moz-wptsync-bot
Harald Alvestrand
Kent Tamura
Robert O'Callahan
Dorel Luca
David Heiberg
Andrea Marchesini
Cameron McCormack
Bogdan Tara
Andrew Swan
Gijs Kruitbosch
Byron Campen [:bwc]
Henrik Skupin
Thomas Wisniewski
Anny Gakhokidze
Brindusan Cristian
James Graham
Luke Bjerring
Rune Lillesveen
kaixinjxq
Kenichi Ishibashi
Ovidio Henriquez
L. David Baron
Yutaka Hirano
Philipp Hancke
jimevans