Repository navigation
Yes, UB is actually that Bad (The Day GCC Broke Chrome) #521
Description
https://gaultier.github.io/blog/the_production_bug_that_made_me_care_about_undefined_behavior.html
Note that ubsan doesn't catch this:
https://compiler-explorer.com/z/z8jx73ncf
And the classic:
Activity
lefticus commentedon Jan 6, 2026
[-]Yes, UB is actually that Bad[/-][+]Yes, UB is actually that Bad (The Day GCC Broke Chrome)[/+]lefticus commentedon Jan 20, 2026
Coming in Ep520
weichslgartner commentedon Feb 16, 2026
Why not mention the memory sanitizer? This is the correct sanitizer for uninitialized memory and detects it in this case: https://compiler-explorer.com/z/4616qzKG6
johnmcfarlane commentedon Feb 18, 2026
@weichslgartner
That's a great question, but worth noting that ASan and MSan don't work together
-fsanitize=memory,addressyieldswhich is a big usability problem!
weichslgartner commentedon Feb 18, 2026
You can use some matrix build in CI to test with all the different sanitizers in parallel.
The more annoying part of MSan is that you need to instrument libc++ (see https://github.com/google/sanitizers/wiki/MemorySanitizerLibcxxHowTo#instrumented-libc)
johnmcfarlane commentedon Feb 18, 2026
Does your matrix include fuzzing, TSan, and axes for toolchains provider, compiler version, build type, etc.? Some warnings and sanitizer features only work with certain flags and compiler releases (as seen in the video). It's gratifying that there is so much activity in this area, but it's not a great ecosystem for your average user. Annoying indeed!😅