Created attachment 16087 [details]
Log file from Pan

Another one from CVS of 15th May

Created attachment 16631 [details]
Pan Log file

This assertion is only in one place:
pan_file_swap_datafile (filename_tmp, filename)
that tries to move the tmp file to the real one.

This is called from all over, in file-grouplist.c,
newsrc-port.c,  filter-xml.c, identity-xml.c, and rule-xml.c,
so it'll take some looking through to find the problem.

If this is still happening for you, Dave, it might be easier
for you to run Pan in gdb, set a breakpoint for log_error_breakpoint,
and wait for the error to show up again, and attach a backtrace.
Would this be possible?

Yes, it's still happening (but not as frequently as before I think).

I'll get you a backtrace.

Dave

Created attachment 18697 [details]
partial backtrace. note concurrent calls to save_groups_to_file().

Add'l info: the above backtrace was generated by getting new headers 
for all subscribed groups.

Chris: I wonder how important it is, actually, to
invoke server_save_grouplist_if_dirty() from task_headers.
The only thing it does here is update the article counts.

I think you're right.  Note that task-headers unrefs the articles just
after that call, which also calls server_save_grouplist_if_dirty (but
protected by a mutex, preventing the problem).

Fixed in CVS:
http://cvs.gnome.org/bonsai/cvsview2.cgi?diff_mode=context&whitespace_mode=show&subdir=pan/pan&command=DIFF_FRAMESET&file=task-headers.c&rev1=1.148&rev2=1.149&root=/cvs/gnome