This is very ad-hoc: I just disabled all tests that failed in three runs on some of our Windows bots. Is there some general pattern that matches the tests I'm disabling, and do other tests match that pattern too? Put differently: Are there other similar tests in this file I should disable?
Diff Detail
Diff Detail
Event Timeline
Comment Actions
I'm able to repro. This bug is happening when the memory pressure on the system is too high (not enough virtual memory).
[ RUN ] SanitizerCommon.SizeClassAllocator64GetBlockBegin ==551160==ERROR: SanitizerTool failed to allocate 0x10000 (65536) bytes of memory at address 0xd74046000 (error code: 1455) ==551160==Dumping process modules: 0x000076fd0000-0x0000770ef000 C:\Windows\system32\kernel32.dll 0x0000771f0000-0x00007739a000 C:\Windows\SYSTEM32\ntdll.dll 0x00013f930000-0x00014081e000 C:\src\llvm\ninja64\projects\compiler-rt\lib\sanitizer_common\tests\Sanitizer-x86_64-Test.exe 0x07fef51d0000-0x07fef52f5000 C:\Windows\system32\dbghelp.dll 0x07fefa620000-0x07fefa623000 C:\Windows\system32\api-ms-win-core-synch-l1-2-0.DLL 0x07fefd060000-0x07fefd0ca000 C:\Windows\system32\KERNELBASE.dll 0x07fefd330000-0x07fefd40b000 C:\Windows\system32\ADVAPI32.dll 0x07fefdc90000-0x07fefdcaf000 C:\Windows\SYSTEM32\sechost.dll 0x07fefe110000-0x07fefe23d000 C:\Windows\system32\RPCRT4.dll 0x07feff330000-0x07feff3cf000 C:\Windows\system32\msvcrt.dll ==551160==Sanitizer CHECK failed: C:\src\llvm\llvm\projects\compiler-rt\lib\sanitizer_common\sanitizer_common.cc:120 ((0 && "unable to mmap")) != (0) (0, 0
To put pressure, I'm using sysinternal "testlimit64".
C:\src\SysInternals>testlimit64 -d -c 18000 Testlimit v5.24 - test Windows limits Copyright (C) 2012-2015 Mark Russinovich Sysinternals - www.sysinternals.com Process ID: 623200 Leaking private bytes with touch (MB)... Leaked 18000 MB of private memory (18000 MB total leaked). Lasterror: 0 The operation completed successfully.
lib/sanitizer_common/tests/sanitizer_allocator_test.cc | ||
---|---|---|
581 | That one will be flaky. It's the same test than above, but with a "fixed" memory address. |
That one will be flaky.
It's the same test than above, but with a "fixed" memory address.
So, issues with memory pressure should be the same.