Simone Atzeni

Hey @SumedhArani, I just forked the "compiler-rt" from the llvm-mirror. I sent you an invitation as a collaborator on that repository, let me know if you are able to read...

@SumedhArani @dongahn Should we close this on Archer since we have everything in the new repo now?

Sometimes are only these: races/critical-unrelated.c races/lock-nested-unrelated.c races/task-dependency.c races/task-taskwait-nested.c and sometimes are more: races/critical-unrelated.c races/lock-nested-unrelated.c races/lock-unrelated.c races/parallel-simple.c races/task-dependency.c races/task-taskwait-nested.c races/task-two.c

@jprotze Did you give a try to this?

@jprotze Maybe we can just create a new wrapper for Fortran and call it `flang-archer` or something like that?

I tried it, it seems better but it can't identify all the races in the task tests, and still a lot of false positives. Here's the list: Failing Tests (19):...

@dongahn @jprotze I think I solved the problem of overriding the weak symbol of `__tsan_on_report` and actually do the reduction without modifying the `compiler-rt`. I created a new branch called...

@dongahn: __tsan_on_report is already weak on tsan, but seems that weak symbols don't work in the same way with static library, that's why we need this workaround.

That's what I know too and I tried to change the order, it didn't work.

@jprotze I see, how would you do that? With dyninst?