[Bug sanitizer/78708] [7 Regression][ASAN][LTO] ICE in expand_ASAN_MARK, at internal-fn.c:380 - when compiling but not linking with ASAN

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=78708

--- Comment #2 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
Works?  You mean it doesn't ICE?  As soon as there are any libasan routine
calls, it can't really link properly, at least not unless you g++ test.o -lasan

Anyway, this is related to PR69254, probably when streaming in LTO bytecode and
seeing there ASAN_MARK internal function with flag_sanitize & SANITIZE_ADDRESS
not set we should set it.  Similarly when streaming in the various UBSAN_*
internal functions, so we don't ICE because sanopt pass has not been invoked.
Of course, the result will not really link, but that is user fault.