Hi list, I'm having a weird issue with the address sanitizer on gcc 10.2.0 (Ubuntu 10.2.0-13ubuntu1), Ubuntu x86_64. I'm running a test bench of code, where the test master forks and executes the individual testcase in the child process. If the child terminates with error (e.g., assertion failed, SIGSEGV or address sanitizer issue), the parent records this as a failure. This setup is integrated using CI. When I pushed some code that actually had an OBVIOUS memory leak, the CI system detected the issue correctly, but locally I cannot reproduce it at all. Unless I compile my code using -O0, it never finds the memory leak. At -O1 it just seems like the memory is not leaking, even though malloc() is called and returns a valid pointer (which is also used, i.e., the memory *is* allocated). As soon as I move the testcase into the parent process (i.e., *same* code!), the address sanitizer correctly detects and reports the issue. I'm compiling my code with -std=c11 -D_POSIX_C_SOURCE=200809L -D_XOPEN_SOURCE=500 -D_DEFAULT_SOURCE -Wall -Werror -Wmissing-prototypes -Wstrict-prototypes -Werror=implicit-function-declaration -Wno-stringop-truncation -O3 -g3 -pie -fPIE -fsanitize=address -fsanitize=undefined -fsanitize=leak I could just fix the obvious leak and move on. But I fear I might not understand why this is happening and have an unreliable CI system that possibly misses leaks. Therefore I would like to understand what is going on here. Any guesses? Thanks, Joe
