Ideally we should wait until the cache is settled, and avoid non-determinism in test cases.
But at the same time, there is no other way to test both conditions, and users might actually experience both cases in the wild.

Tests are in D271031

Tests are in D271027 .

Tests are in D271025

This will always iterate over all memory-cached script of the client processes.
I guess we can keep it as-is for now and revisit if this becomes a problem.

I was surprised to not see as many code changes in the JITs, but these are already covered by the EnvironmentCoordinate which are making use of function modified in this patch.

🎉

Good catch!

The number of MOZILLA_INTERNAL_API is quite unfortunate.
I am tempted to suggest to create a public directory to hold the non-internal classes separately, but this will not work either as the signature of methods imply having definitions of the internal types.

Good catch for the missing DropJSObjects.

nit: s/Part 6/Part 7/

I am slightly troubled by the fact that we can have a mStencil set without updating the mDataType.
Maybe we should update the mDataType comment to reflect the status of mStencil relative to the data type.

note: the naming is a little strange, but it makes sense with the part 2.

This revision requires a Testing Policy Project Tag to be set before landing. Please apply one of testing-approved, testing-exception-unchanged, testing-exception-ui, testing-exception-elsewhere, testing-exception-other. Tip: this Firefox add-on makes it easy!

This revision requires a Testing Policy Project Tag to be set before landing. Please apply one of testing-approved, testing-exception-unchanged, testing-exception-ui, testing-exception-elsewhere, testing-exception-other. Tip: this Firefox add-on makes it easy!

Nice!

nit: vixl::Assembler::ShiftDP and vixl::Assembler::ExtendMode should cast the enumerated values back to int32_t before doing a left shift.

This revision requires a Testing Policy Project Tag to be set before landing. Please apply one of testing-approved, testing-exception-unchanged, testing-exception-ui, testing-exception-elsewhere, testing-exception-other. Tip: this Firefox add-on makes it easy!

Try run: https://treeherder.mozilla.org/jobs?repo=try&revision=cd4e6cabd2cdd5b751c170efa1c3163785fe3726&selectedTaskRun=bx5Deoq1Tw6w3_MSlkcWwA.0
As mentioned in the bug, this seems to have been fixed by Bug 1977692.
Thus, only landing the test case modification to avoid regressions.

Looking back at the results, the previously reported regressions on jetstream 3 are now gone with extra runs.

Sorry I had not seen Bug 1981918

Sorry, I had not seen Bug 1981921

Given the table at https://glandium.github.io/firefox-linux-compat-matrix/#scope=build&version=136&show_eol=false
We are still looking to support clang 8.0 or above.

Given the table at https://glandium.github.io/firefox-linux-compat-matrix/#scope=build&version=136&show_eol=false
We are still looking to support GCC 8.1 or above.

Re-running the benchmark using a proper base revision: https://perf.compare/compare-lando-results?baseLando=149241&newLando=149242&baseRepo=try&newRepo=try&framework=13
There is some jetstream 3 regression, I am spinning extra iterations to have more confidence.