I cannot find a culprit CL. This could be an infra issue...?

I think it is quite likely an infra issue. Tests failing to map a file from disk and there are no relevant changes in the regression range.

First failure is at https://build.chromium.org/p/chromium.webkit/builders/WebKit%20Android%20%28Nexus4%29/builds/70086 and this is immediately following another suspicious (but not the same) failure where device provisioning failed: https://uberchromegw.corp.google.com/i/chromium.webkit/builders/WebKit%20Android%20%28Nexus4%29/builds/70085

ack

I think the first failing build is https://build.chromium.org/p/chromium.webkit/builders/WebKit%20Android%20%28Nexus4%29/builds/70085

The provision_devices failure is a red herring. Network flake interrupted the slave's connection to the master. Had the build gotten to the point of running tests, webkit_unit_tests might have failed.

In fact, in that build's blamelist is the cl https://chromium-review.googlesource.com/c/chromium/src/+/677449 which looks pretty relevant. Could that be responsible?

Seems quite plausible. I'll revert the CL and we'll see.

Seems to be more than one issue with webkit_unit_tests. Jumbo (though disabled) seems to have added a v8 startup crash (bad compiler flags?) which the revert fixed but there is now other failures and it is hard to say exactly when they were introduced since the test crashed for several hours. Unless there is another bot which didn't crash?

70083 Scrolling test failure
70084 Successful!
70085 Exception provison devices (seems to have lost contact with the device)
70086 webkit_unit_test: Failed to list tests on device, startup crash v8
...
70132 webkit_unit_test: Failed to list tests on device, startup crash v8
jumbo revert landed 15:35
70133 webkit_unit_tests ImageDocumentTest.ImageCenteredAtDeviceScaleFactor
70134 webkit_unit_tests ImageDocumentTest.ImageCenteredAtDeviceScaleFactor
....

Yes, the now fixed mmap issue masked the ImageDocumentTest.ImageCenteredAtDeviceScaleFactor so it's hard to find the culprit of that in the build list. I'll file a separate bug for it.

Filed  issue 772860  (and found likely culprit CL).

And I've found the cause of the android problem. Bad packaging in android because I incorrectly split a test() into test()+source_set().

So we're done here. Thanks for reverting it dpranke!