[cifuzz] Don't enforce OOM limit when no OOMs aren't reported

[jonathanmetzman]

Fixes google/clusterfuzzlite#66

[evverx]

@jonathanmetzman thanks! I think by analogy with this, timeout shouldn't be set when REPORT_TIMEOUTS is set to false. Looking at

artifact_prefix='/tmp/tmp_ez5oezd/'; Test unit written to /tmp/tmp_ez5oezd/timeout-e901ba1ebbf2a7329356892381b268574dca574a
==21== ERROR: libFuzzer: timeout after 25 seconds
    #0 0x52f411 in __sanitizer_print_stack_trace /src/llvm-project/compiler-rt/lib/asan/asan_stack.cpp:87:3
    #1 0x470a38 in fuzzer::PrintStackTrace() cxa_noexception.cpp
    #2 0x4557c9 in fuzzer::Fuzzer::AlarmCallback() cxa_noexception.cpp

it seems timeouts are still triggered. (in this particular case it's an inifinite loop that is supposed to be fixed in a patch I'm testing but slow code nobody cares about triggers timeouts much more often I think)

One last question. CFlite is pinned to v1 in my GHAction. Will it work automagically once this PR is merged or should I point it to master?

[jonathanmetzman]

@jonathanmetzman thanks! I think by analogy with this, timeout shouldn't be set when REPORT_TIMEOUTS is set to false. Looking at

artifact_prefix='/tmp/tmp_ez5oezd/'; Test unit written to /tmp/tmp_ez5oezd/timeout-e901ba1ebbf2a7329356892381b268574dca574a
==21== ERROR: libFuzzer: timeout after 25 seconds
    #0 0x52f411 in __sanitizer_print_stack_trace /src/llvm-project/compiler-rt/lib/asan/asan_stack.cpp:87:3
    #1 0x470a38 in fuzzer::PrintStackTrace() cxa_noexception.cpp
    #2 0x4557c9 in fuzzer::Fuzzer::AlarmCallback() cxa_noexception.cpp

it seems timeouts are still triggered. (in this particular case it's an inifinite loop that is supposed to be fixed in a patch I'm testing but slow code nobody cares about triggers timeouts much more often I think)

Maybe, it might be a good idea to enforce timeouts even if they are not reported so we can run other fuzzers. Need to think about this.

One last question. CFlite is pinned to v1 in my GHAction. Will it work automagically once this PR is merged or should I point it to master?

Yes.

[evverx]

it might be a good idea to enforce timeouts even if they are not reported so we can run other fuzzers

Agreed. But I think it should be "total_time / the number of fuzz targets" instead of 25s when REPORT_TIMEOUTS is set to false