Closed nbriggs closed 4 years ago
For the varinfo/origins and leaks/gdbserver testcases, at least we have some idea of what is going wrong.
fprw just looks like a codegen difference causing different types of error.
I haven't analyzed the others. annotate_trace_memory has a difference in the order of the traces.
The OpenMP testcase failures are either because we're missing some suppressions or else there is a missing libomp redirection.
For comparison on amd64 I get the following extra failures
< memcheck/tests/amd64/insn-pmovmskb (stderr) < memcheck/tests/clientperm (stderr) < memcheck/tests/signal2 (stdout) < memcheck/tests/signal2 (stderr) < helgrind/tests/tc22_exit_w_lock (stderr) < drd/tests/atomic_var (stderr) < drd/tests/tc04_free_lock (stderr) < none/tests/amd64/ssse3_misaligned (stderr)
But these one pass on amd64
gdbserver_tests/mcblocklistsearch memcheck/tests/leak-tree memcheck/tests/leak_cpp_interior memcheck/tests/varinfo5 drd/tests/annotate_trace_memory drd/tests/annotate_trace_memory_xml
I've moved this info to the wiki page.
This is more informational than a report of a specific problem. Using valgrind (valgrind-3.17.0.GIT-39c9c1ec89-20200621X) compiled with clang 9.0.1 running the regtest compiled either with the same clang, or with gcc ((FreeBSD Ports Collection) 9.3.0) shows up the following differences in test failures:
clang0620.txt gcc0620.txt