Jakub Jelinek wrote:
> The problem is that at least by default we
> don't want very expensive tests in the testsuite, whether it is runtime
> expensive or eats too much memory.
I see. We can only merge small tests then. From what I see in
compiler-rt/lib/tsan/lit_tests, most of the tests are rat
Konstantin Serebryany wrote:
> Other mode may work too, but we will not own the tests
> (similarly, we don't own the existing asan tests in GCC,
> although if the tests need fixing during integrate we do it).
I think everyone agrees on this. Once tests are in, Gcc team will
maintain them.
-Y
On Mon, Dec 02, 2013 at 02:13:42PM +0400, Yury Gribov wrote:
> Currently gcc has tests for AddressSanitizer and UBSanitizer. Do you
> think it would make sense to add support for ThreadSanitizer testing
> as well? If the answer is positive, we can work on dg infrastructure
> (tsan-dg.exp, tsan.exp)
I'd be glad to have tsan tests in GCC.
At the very least we need to have a couple of sanity tests to make
sure tsan links and finds a trivial race.
The only mode in which my team can truly support tsan tests is when
they are verbatim copies of the upstream tests
and they are merged together with th
Hi all!
Currently gcc has tests for AddressSanitizer and UBSanitizer. Do you
think it would make sense to add support for ThreadSanitizer testing as
well? If the answer is positive, we can work on dg infrastructure
(tsan-dg.exp, tsan.exp) and initial set of tests. As for the latter:
should we
