Hi (adding Philippe to CC who knows a bit more about Valgrind vs Ada), On Sat, Oct 05, 2024 at 06:07:27AM +0100, Sam James wrote: > Jeff Law <jeffreya...@gmail.com> writes: > > > On 10/2/24 8:39 PM, Sam James wrote: > >> Valgrind doesn't error out by default which means bootstrap issues like > >> in PR116945 can easily be missed: pass --exit-errorcode=1 to handle this. > >> While here, also set --trace-children=yes to cover child processes > >> of tools invoked during the build. > >> Note that this only handles tools invoke during the build, it > >> doesn't > >> cover everything that --enable-checking=valgrind does. > >> gcc/ChangeLog: > >> PR other/116945 > >> PR other/116947 > >> * configure: Regenerate. > >> * configure.ac (valgrind_cmd): Pass additional options. > > But is this going to cause all bootstraps with Ada to fail? That's > > how I read 116945 which was closed as WONTFIX. Or am I > > mis-interpreting that BZ and its interaction with this patch? > > No, you're right, I consider this on pause unless/until we figure out > that bug -- I'm speaking with mjw about some ideas.
The problem as explained in the bug https://gcc.gnu.org/bugzilla/show_bug.cgi?id=116945 is that we don't understand what code is generated for the guard that checks whether the value is undefined. I assume the 'Valid guard relies at least on (partially) defined bits (but maybe it really is not deterministic?) Depending on what the generated code actually looks like we can either try to get memcheck more accurate, add valgrind annotations like in ggc (but do we have an Ada variant of the valgrind.h macros?) or add a suppression (where another wrinkle is that Valgrind doesn't seem to demangle Ada symbols, https://bugs.kde.org/show_bug.cgi?id=445235). Cheers, Mark
