Re: [AC21.5] segfault during compilation

HN> This change will probably not fix any bug. In C, '\0' is another HN> way to spell 0, and in C++ '\0' will be cast to 0 anyway, since HN> memset's second argument is int. HN> Not that I'm against the change -- using '\0' makes the HN> *intention* a bit clearer -- but it should be applied with HN> understanding that it doesn't resolve any real problem. Right, this part of the patch does not address the problem. I just stumbled across this memset call and adjusted it to the others. HN>> + memset (&mc_allocator_globals, '\0', sizeof HN>> (mc_allocator_globals_type)); HN> This shouldn't be necessary unless init_mc_allocator is called HN> twice (which it shouldn't be, judging by the comments and the HN> code). A global variable is 0-initialized by default and calling HN> memset doesn't accomplish anything. An uninitialized lookup table is the only thing I can currently think of that may cause the reported failure. Although, I was quite surprised that this bug didn't hit earlier and more often. I'll keep on investigating. Do you want me to revert this change? -- Marcus