Here is a problem which has long troubled me with XEmacs and is still present in 21.5. Subprocesses which produce a lot of output cause XEmacs to miss interrupt signals, or at least, take a very long time to react to them and pass to the subprocess. Compare this test in XEmacs 21.5 (or earlier) and any version of GNU Emacs: M-x shell yes RET (the command "yes" prints y indefinitely to the terminal). Wait 30 secs or so and try to interrupt the XEmacs shell with C-c C-c. The signal doesn't easily get through. If, however, you hit C-c C-c soon after "yes" has started it is interrupted successfully. There is no problem in GNU Emacs, the signal gets passed quickly to the subprocess. I'm using Fedora Core 4/i386 but I think the problem exists on other Unices, at least. This is a real problem when writing code which communicates with subprocesses that may go awry producing output indefinitely. - David.