Open nickdesaulniers opened 1 year ago
@llvm/issue-subscribers-debuginfo
What would be your preferred source location for the stack check setup -- presumably the same as the prologue?
Putting a source location on the call to __stack_chk_fail
is slightly more tricky seeing how multiple return paths presumably branch to the call. Possibly the closing brace of the function would be most appropriate, however I'm not sure if that's available so late in compilation.
What would be your preferred source location for the stack check setup -- presumably the same as the prologue?
SGTM
Perhaps the epilog can be reused for __stack_chk_fail if available?
Also, it looks like -g1
changes the debug info associated? (The coverage looks better with -g1
than -g2
).
Consider the following:
It looks like the following assembler is lacking debug info:
This is more easily visible in Godbolt: https://godbolt.org/z/cnozfa15E