"Stephen J. Turnbull" <turnbull(a)sk.tsukuba.ac.jp> writes:
I've discovered that in this XEmacs, if a file is truncated due
to
lack of disk space on write, XEmacs blithely goes ahead and renames
the tmp file and does whatever it usually does (including remove) to
the previous version.
Eek. We had this problem several years ago, it was fixed, and now
it's recurring.
When the write fails, do you see any error whatsoever? If no, could
you try setting `debug-on-signal' to t? Then all errors signaled
(even those that are trapped) will throw you into the debugger, and
you can press `c' to proceed to the next error.
In case you feel like investigating it on your own, the error should
be detected in the streams code, either in the writer or in the
destructor (because of the fflush()) of the stream. Our bug of
several years ago was in the latter case, where the error was not
properly propagated to the code above.