On Friday 25 October 2002 06:07, you wrote:
> I have a frustrating segfault problem that I can't seem to track down.

It *may* have nothing to do with your problem, but beware of
"clock scan ... �gmt 1" in your scripts.

The problem is in the "-gmt 1". This is broken in Tcl (not AOLserver)
on Sun platforms. Not broken in terms of functionality but in terms
of MT-safety. The environment array is getting corrupted by unprotected
modification of the TZ variable. This leads to all sorts of memory
problems, including one you've described.
This was one of the *most* difficult problem I've tracked down in the
AOLserver environments.

Also, if some of your code is (indirectly) accessing the process environment
(like tzset() or similar) chances to get hit by this bug are even greater.

I have submitted a bug report on Tcl project for that.

Again, it may have nothing to do with your particular problem, but it's
worth investigating in this direction. Of course, if you have some strange
Purify reports I'd attend them first. Be sure to recompile the entire
suite (AOlserver, TCL, etc) so that internal object allocation/deallocation
and generally all memory reservation calls (alloc/free) are mapped to
system malloc/free, since this way Purify works the best.

Watch also for *hidden* usage of non-mt-safe functions in your C-code.
Often, people neglect this fact, which leads to very difficult, hard
to trace problems.

Cheers
Zoran

Reply via email to