Re: [FAILURE] XEmacs 21.0-beta38 on MkLinux

Stephen> hniksic(a)srce.hr said: Stephen> Don't blame Fred...at least he found OUR problem. He didn't claim to have Stephen> found the correct solution from the larger XEmacs point of view. As he says in Stephen> a later note to me, The source will soon have this comment: /* In C++, it is impossible to determine what type matherr expects without some more configure magic. We shouldn't be using matherr anyways - it's a non-standard SYSVism. */ The unilateral change to __exception must be a mistake in glibc. It will likely be fixed in any released version of glibc. Someone who understands floating point error handling should look at this and likely expunge it. My guess is that a signal handler for SIGFPE is all a modern system needs to have. FLOAT_CATCH_SIGILL should be nuked too. Martin