On Thu, Mar 05, 2026 at 06:00:44PM +0000, Stuart Henderson wrote:
> On 2026/02/25 15:29, Brad Smith wrote:
> > ping.
> 
> make test takes flipping ages on this!
> 
> amd64 (IBT machine) was ok up to "test shard 8 of 10" and then hit this
> (it stops running after a shard with a failure):
> 
> [----------] Global test environment tear-down
> [==========] 975 tests from 54 test suites ran. (1180357 ms total)
> [  PASSED  ] 974 tests.
> [  FAILED  ] 1 test, listed below:
> [  FAILED  ] VP9/RealtimeTest.IntegerOverflowLarge/0, where GetParam() = (0xd98b238c6d8, 0)
> 
> suspect this is not really a problem, and would have hit any IBT issues
> long before if they were any.
> 
> aarch64 still running on my M2 (it's partway through 8 of 10, no
> failures yet). will update if there are any failures otherwise assume
> it was ok. again any BTI issues would have happened long before.
> 
> we also need to be sure that it doesn't break chromiums (including
> qtwebengines which are on earlier chromium codebases)

That test also fails in current in 1.15.2, but it succeeded after a
ulimit -Sd $((22 * 1024 * 1024)) (yes, 22GB--I can confirm that 20GB
isn't enough). Same applies for 1.16.0, and I did test on IBT hardware,
with no failures.