Ian T Zimmerman <itz(a)rahul.net&gt; writes: Yes I do. However we have been seeing these bug reports for quite some time now on XEmacs. I guess even before the big move towards glibc2. So there must be something else happening too (or my recollection is wrong/all these bug reports were from glibc early adopters). So like I commented the code was written for a different signal behaviour and you think this changed during the libc5->libc6 change over. I am fairly sure some versions ago XEmacs used those flags on Linux. Could it be that for some reason using this flag in the application breaks the library too? I am not at my linux system so I cannot check what the mechanism is. I am not sure but I am fairly certain that on every glibc6 system (and there are quite a few now) the signal() behaviour is now as I described and therefore libgpm is broken there. I was confused in thinking this was universal because the Solaris 2.x 'signal' manpage also decribes the 'bsd' behaviour. All the bug reports we had were from people running stable Kernels. So at the very least you would have to deal with 2.0.x+glibc2. On my glibc2 system this is happening for sure. Ok thanks. It will kill off a lot of bug reports in the future. Note that you already special case SIG_IGN when installing the handler. I was being confused by both the Solaris and the glibc2 docs (or at least the man paged derived from it for the Debian distribution) describing 'signal' as a BSD compatibility layer over sigaction. Thank you very much. Jan