Test failures building RELEASE_3.9.0/final

I’ve “successfully” built 3.9.0 release but when I run “ninja check-all” I got 208 Unexpected failures:

Expected Passes : 33997
Expected Failures : 198
Unsupported Tests : 685
Unexpected Failures: 208

Below is the log I captured running “time ninja check-all | tee ninja-check-all.txt”


Suggestions on next steps?

– Wink

These all look related to some kind of sanitizer-specific crash on startup. If you’re not using sanitizers, maybe you don’t care about these failures. If you want to file bugs and get them fixed, try running ‘check-asan’ and ‘check-msan’. Extract the failing command from a test case, run it under a debugger, and file a bug with the crash stack trace. There is probably some environmental issue on your system that makes shadow memory allocation fail, or causes an early shadow memory access.

Almost all of the libc++ failures are due to GCC bugs. I’ll work on silencing the failures but the libc++ ones are nothing to be concerned about.

I looked at the buildbots but I didn’t archives for release builds. I was thinking I could compare my results with a buildbot version. Are there any archives for 3.9.0?

I ran “ninja check-asan” and no errors. But “ninja check-msan” had 117 errors.

I took the first FAILED test, which was for eventfd.cc, and executed the command
line creating an eventfd executable in a temporary directory and then
executed that file using gdb. Finally, used bt to dump the stack.

I’ve emailed llvm-admin@lists.llvm.org to setup an account since
self-registration is disable. After getting an account I’ll then file a bug.
If there’s anything else anyone thinks I should do let me know.

Here’s the output from the FAILED test:

– Testing: 120 tests, 12 threads –
FAIL: MemorySanitizer-x86_64 :: Linux/eventfd.cc (1 of 120)
******************** TEST ‘MemorySanitizer-x86_64 :: Linux/eventfd.cc’ FAILED ********************

Never seen anything like this. What's the platform?

Arch Linux see a little more info here