I think the following test case demonstrates a valgrind bug, but it is a strange corner case, so maybe it's just a case that requires a suppression. Specifically, the following error is unexpected, and happens for SIGSEGV, but not for SIGUSR1. ==19965== Invalid read of size 4 ==19965== at 0x8048475: test (in /home/vandebo/valgrind/jmptest) ==19965== by 0x8048547: main (in /home/vandebo/valgrind/jmptest) ==19965== Address 0xFEFFD3BC is on thread 1's stack Gcc will sometime generate the short assembly segment in the test case with -Os, but is dependant on the version of gcc and the surrounding code. -- Steve |