misc. annoyances

[Scott McKay <SWM@sapsucker.scrc.symbolics.com>]

    Date: Tue, 2 Jun 87 11:30:36 pdt
    From: king@kestrel.ARPA (Dick King)

    1> Does anyone have an "autosave" package?  Autosave is a feature
       whereby dirty file buffers are periodically written to a file whose
       name is unusual but can be determined from the name of the buffer's
       file.  A refinement of such a package would be to delete the
       autosave file when the buffer is written, thereby avoiding file
       system clutter.

    2> Well, the editor isn't reentrant.  Again.  Could ZWEI compilation
       and evaluation take place in a separate process?  Forms rather than
       strings could be sent, reducing problems with packages, etc.
       *STANDARD-OUTPUT* in this process could be bound to a stream that
       arranges for the ZWEI process to print out routine output, but
       *DEBUG-IO* to the process' Background Interactor.  If the
       evaluation process took a break then ZWEI, which would be running
       (say) COM-SEND-FORMS-TO-ZWEI-BACKGROUND-AND-AWAIT-COMPLETION, could
       be aborted.

       The minor annoyance of shifting to the background interaction
       window for breaks or aborts is far outweighed by the avoided loss
       of C-E and other use of the editor.

There will be multiple Zmacs processes in 7.2, complete with a robust
locking scheme on critical datastructures to prevent two Zmacs processes
from shafting each other.

    3> I don't know if this is the right place to report this bug, but
       I'll do it anyway.  The patch system has no interlocking at all.  A
       while ago a co-worker and I were assigned the same patch number in
       the obvious
	  I-read-patch-directory  --- he-reads-patch-directory   ---
	  I-write-patch-directory --- he-writes-patch-directory 
       sequence.  The fix is to check that you got the proper version
       number when you open the patch directory for output to write a
       modified version, and to start over completely if you don't.
       Writing a UNIX-style lock file seems wrong here.

There is in fact interlocking on patch-directories.  However, there is a
small but significant window (which has proved very difficult to close
completely) in which two people can bash each other.  In practice, this
turns up often enough to be annoying, but not so often as to be
completely awful.  The window increases in size depending on how slow
the file server is.

    Thanks for reading...

    -dk