🤖 New build scheduled with the buildbot fleet by @ZeroIntensity for commit 3d22310 🤖

Results will be shown at:

https://buildbot.python.org/all/#/grid?branch=refs%2Fpull%2F133040%2Fmerge

If you want to schedule another build, you need to add the 🔨 test-with-buildbots label again.

@picnixz This seems to have resulted in the -ldl argument being included twice on macOS and iOS builds. It's not causing an error, but it is raising a lot of compiler warnings

Looks like it's not just macOS and iOS; Ubuntu builds now include compilation statements like gcc -shared -ldl -ldl -ldl -ldl -ldl -ldl .... It doesn't look like GCC is complaining... but I think dl might be linked. 🤣

Aaaaaah I don't understand:( Why would it include it twice I'll investigate

Instead of adding it to all LDLFLAGS, I'll only add it the LDLFLAGS for faulthandler

Mmh, it's actually in Python/traceback.c. A bit more annoying.

Hum. I also forgot removing defined(HAVE_BACKTRACE_SYMBOLS) after Victor told me that backtrace_symbols wasn't used.

faulthandler is built as a built-in module, since Py_FatalError() uses it (dump the backtrace and than turns off faulthandler).

I found the issue. Because of AC_CHECK_HEADERS + AC_CHECK_FUNCS, we actually expand LDFLAGS 2x3=6 times, hence the 6 -ldl -ldl -ldl -ldl -ldl -ldl.

PR is ready: #133071.

It doesn't look like GCC is complaining... but I think dl might be linked. 🤣

And yes, that's why I didn't see the warnings. I use gcc and not clang

I was wondering if autoreconf was smart enough to deal with the extra flags. Buildbots passed so I figured it was ok :(.