Closed mborland closed 3 months ago
Both of these options should already be present. -g
is supplied by <debug-symbols>on
:
https://github.com/bfgroup/b2/blob/3c0f5630157c3db1025b54c2f6da25cf234b8e20/src/tools/gcc.jam#L678
and -fno-omit-frame-pointer
is supplied by <address-sanitizer>norecover
:
https://github.com/bfgroup/b2/blob/3c0f5630157c3db1025b54c2f6da25cf234b8e20/src/tools/gcc.jam#L695
The features, in turn, are supplied by drone.sh when ASAN is set:
Thanks as always @pdimov. I was hoping to get more since these failures seem extremely odd like limits_link_test having a seg fault on Drone while the fuzzers on GHA are fine. I'll try adding the same platform ASAN to GHA.
The failures in #209 are a bit weird because (a) everything fails and (b) the push Drone job (which should be testing the same thing) succceeds.
The failures in #209 are a bit weird because (a) everything fails and (b) the push Drone job (which should be testing the same thing) succceeds.
Yeah the ASAN GCC-12 runs fine on GHA. @sdarwin could this possibly be related to the image changes you made on Drone yesterday? The push job ran 4 days ago.
I'm seeing the odd segmentation faults in decimal as well: https://drone.cpp.al/cppalliance/decimal/1092/40/2. Many of these test are purely math operations.
PR 211 succeeded in Drone
Drone log from PR #209 shows asan failure without any debug information or stacktrace.