The stacks I looked at look like:
==936==ERROR: LeakSanitizer: detected memory leaks
Direct leak of 25568 byte(s) in 1 object(s) allocated from:
    #0 0xcdb522 in __interceptor_realloc /b/build/slave/linux_upload_clang/build/src/third_party/llvm/compiler-rt/lib/asan/asan_malloc_linux.cc:117:3
    #1 0x7fcd1ec4c952  (/usr/lib/nvidia-384/libEGL_nvidia.so.0+0x33952)

We probably can't do anything about leaks in the driver; maybe we should suppress things with libEGL_nvidia on the stack? (That'd also suppress things caused by chromium code leaking gl resources I suppose.)

A little surprised that "EGL" is in the library name since Chromium's Linux port uses GLX rather than EGL, at least by default. On the whole though I think that suppressing these reports and getting the tests running is better than not running the tests at all because of these leak reports. piman, what do you think?

This is testing a configuration that we don't support (EGL on Linux), but was added by third-party, and tests code that is used by Chrome OS, so I agree that keeping the tests in is valuable. I don't have the bandwidth to dig into the failures though. I'm fine with us suppressing libEGL_nvidia things, as we'd never use that in production (on either Linux desktop nor Chrome OS).

It passes now: https://ci.chromium.org/p/chromium/builders/luci.chromium.try/linux_chromium_asan_rel_ng/40371

I wonder what changed...looks like they no longer run on NVIDIA GPU. I wonder what caused that.

Oh, comment 0 was a try run for a try job that runs them on GPU: https://chromium-review.googlesource.com/c/chromium/src/+/1089589/2

thakis: are you saying this bug is fixed now, and should be closed?