Fix tsan/asan/hwasan/etc failures on Android

[charlesmunger]

Some people run their benchmarks with sanitizers, and they should not spuriously fail.

This unblocks some test failures in google3

[LebedevRI]

Could you please adjust the commit message to contain the previous output of each of those lines?

[LebedevRI]

@charlesmunger poke

[LebedevRI]

@charlesmunger poke

[charlesmunger]

I'm not sure what you mean? The asan allocator returns 0 for all mallopt calls regardless of the params. The output is just that the BM_CHECK triggers.

[LebedevRI]

I'm not sure what you mean? The asan allocator returns 0 for all mallopt calls regardless of the params. The output is just that the BM_CHECK triggers.

That's what the commit message should say. I'm just asking you to put the reasoning in the commit message.

[dmah42]

i don't suppose there's any way to check if we're under sanitizer at runtime? it would be great to keep this check when not running under sanitizer.

[LebedevRI]

i don't suppose there's any way to check if we're under sanitizer at runtime? it would be great to keep this check when not running under sanitizer.

Sanitizers are compile-time opt-in, they can be detected the usual C way.
I'm not really sure what is the correct list of sanitizers that override mallopt, but i'd guess something like

#if __has_feature(address_sanitizer) || __has_feature(memory_sanitizer) || __has_feature(thread_sanitizer) || __has_feature(hwaddress_sanitizer)
// no BM_CHECK
#else
// can check return value
#end

[charlesmunger]

I did try that at first, but new sanitizers are being added all the time, like dfsan and typesanitizer.

[LebedevRI]

@charlesmunger thank you!