Also affects other Linux bots (not just ASan).

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/c4c4fd40e57affab6a53322119636a09aed77c55

commit c4c4fd40e57affab6a53322119636a09aed77c55
Author: mgiuca <mgiuca@chromium.org>
Date: Mon Nov 21 02:19:53 2016

Revert of Minor style changes to HarfBuzzShaper to improve readability (patchset #1 id:1 of https://codereview.chromium.org/2513473007/ )

Reason for revert:
Appears to have broken HarfBuzzShaperTest.ResolveCandidateRunsUnicodeVariants on Linux ASan.

BUG= 667147 

Original issue's description:
> Minor style changes to HarfBuzzShaper to improve readability
>
> TBR=szager@chromium.org
>
> Committed: https://crrev.com/95b18a0f83f57c597c435e09b4081d995e0defe0
> Cr-Commit-Position: refs/heads/master@{#433265}

TBR=szager@chromium.org,eae@chromium.org
# Not skipping CQ checks because original CL landed more than 1 days ago.

Review-Url: https://codereview.chromium.org/2517023002
Cr-Commit-Position: refs/heads/master@{#433469}

[modify] https://crrev.com/c4c4fd40e57affab6a53322119636a09aed77c55/third_party/WebKit/Source/platform/fonts/shaping/HarfBuzzShaper.cpp

That revert didn't fix it:
https://build.chromium.org/p/chromium.webkit/builders/WebKit%20Linux%20Trusty/builds/20177

Looking...

The first failing build was:
https://build.chromium.org/p/chromium.memory/builders/Linux%20ASan%20LSan%20Tests%20%281%29/builds/31402

Suspect range:

r433235..r433314 (Linux ASan LSan Tests)
r433363..r433371 (WebKit Linux Trusty)

Well now that's just very confusing (the ranges don't overlap).

Linux ASan LSan Tests range (r433235..r433314):
https://chromium.googlesource.com/chromium/src/+log/7f6421b1628a15833196ff7cf1fefbe5fcd69558..0cd059db8c5ba951d39e37b53113802aa04a2e78?pretty=fuller

WebKit Linux Trusty range (r433363..r433371):
https://chromium.googlesource.com/chromium/src/+log/8f26cf4991ebfaf35bf64d976d66d97b8edba360..2de9572364f3e2b29aad7837ec0d36651464a1f7?pretty=fuller

I looked at the code in the test.

https://cs.chromium.org/chromium/src/third_party/WebKit/Source/platform/fonts/shaping/HarfBuzzShaperTest.cpp?l=95

// If the specified VS is not in the font, it's mapped to .notdef.
// then hb_ot_hide_default_ignorables() swaps it to a space with zero-advance.
// http://lists.freedesktop.org/archives/harfbuzz/2015-May/004888.html
// OpenType recommends Glyph ID 3 for a space; not a hard requirement though.
// https://www.microsoft.com/typography/otspec/recom.htm

This sounds like a bit of a dodgy test just from that description (asserts that space is Glyph ID 3, even though it isn't mandated by the spec). It's possible that a font change has happened across the fleet, but rolled out at slightly different times which explains why a) I can't find a CL that caused this, and b) the breakage is introduced in non-overlapping (but close) windows between two different bots.

Spoke with tansell@. He says chrishall@ was looking at a bug where fonts were changing on bots and they didn't know why.

For now I might disable the test since a culprit CL can't be found.

This logic was introduced in https://crrev.com/48a6aeba789269ef52dd8389f1cd01b24c35fe9d.

Disabling in https://codereview.chromium.org/2516273002 and assigning this bug to kojii@chromium.org.

(If it isn't an important assertion, just remove it I suppose.)

Thank you for looking into this and sorry for the trouble, I'm surprising this is hitting after more than one year.

I'll check the code and will remove it or make it robust.

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/f324f248c00be01a63ada4e40e20ea58ca345666

commit f324f248c00be01a63ada4e40e20ea58ca345666
Author: mgiuca <mgiuca@chromium.org>
Date: Mon Nov 21 06:26:17 2016

Partial-disable HarfBuzzShaperTest.ResolveCandidateRunsUnicodeVariants.

This started failing on Linux bots, in different commit windows
(indicating that a configuration, perhaps font, has changed, not a
regression). Disabling to get the tree green, and someone can
investigate.

BUG= 667147 
TBR=kojii@chromium.org

Review-Url: https://codereview.chromium.org/2516273002
Cr-Commit-Position: refs/heads/master@{#433485}

[modify] https://crrev.com/f324f248c00be01a63ada4e40e20ea58ca345666/third_party/WebKit/Source/platform/fonts/shaping/HarfBuzzShaperTest.cpp

 Issue 667094  has been merged into this issue.

We did change some of the install build deps, and I wonder if that might've affected some of the fonts? 

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/4a4861960ae3a020a3fcdaa417216e7f7411792c

commit 4a4861960ae3a020a3fcdaa417216e7f7411792c
Author: eae <eae@chromium.org>
Date: Tue Nov 22 03:16:24 2016

Reland of Minor style changes to HarfBuzzShaper to improve readability

Landed in r433265, reverted in r433469 due to an unrelated asan failure.

R=szager@chromium.org,mgiuca@chromium.org
BUG= 667147 

Review-Url: https://codereview.chromium.org/2517193002
Cr-Commit-Position: refs/heads/master@{#433768}

[modify] https://crrev.com/4a4861960ae3a020a3fcdaa417216e7f7411792c/third_party/WebKit/Source/platform/fonts/shaping/HarfBuzzShaper.cpp

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/c3d0368b185088d697bad36d0928a7409d1c970a

commit c3d0368b185088d697bad36d0928a7409d1c970a
Author: kojii <kojii@chromium.org>
Date: Tue Nov 22 15:54:26 2016

Make HarfBuzzShaperTest.ResolveCandidateRunsUnicodeVariants more robust

Some unknown changes to Linux bots figured out that this test is not as
robust as it should be, by relying non-mandatory recommendation.

Since it has started failing on Linux bots, it was disabled in [1].
This patch makes it more robust to fonts that do not follow the
OpenType recommendations and re-enable it.

[1] https://codereview.chromium.org/2516273002

BUG= 667147 

Review-Url: https://codereview.chromium.org/2520833002
Cr-Commit-Position: refs/heads/master@{#433874}

[modify] https://crrev.com/c3d0368b185088d697bad36d0928a7409d1c970a/third_party/WebKit/Source/platform/fonts/shaping/HarfBuzzShaperTest.cpp
[modify] https://crrev.com/c3d0368b185088d697bad36d0928a7409d1c970a/third_party/WebKit/Source/platform/fonts/shaping/ShapeResultTestInfo.cpp
[modify] https://crrev.com/c3d0368b185088d697bad36d0928a7409d1c970a/third_party/WebKit/Source/platform/fonts/shaping/ShapeResultTestInfo.h