Date   : Wed, 10 Jun 1981 10:19:00-PDT
From   : fylstra.tsca@Sri-Unix
Subject: bugs in SUBMIT/XSUB

I have noticed the following bugs in SUBMIT and XSUB:

       1. XSUB chokes on empty lines and dies a miserable death.
       I noticed this while trying to spoon feed a number of
       answers to prompts that Small C (Codeworks version) was
       asking.  My intention was to reply (from the .SUB file)
       with a carriage return.

       A systems programmer from Digital Research acknowledged
       this at the recent West Coast Comp Faire.  He said that
       they knew it was a kludge but "it's better than nothing..."

       2. When a transient program returns to the CCP via a
       RET instruction rather than doing a warm boot, the CCP
       fails to restore the DMA address to TBUFF at 80h.
       If a .SUB file is active, the CCP will try to read the
       next command line from the file into some random buffer
       (most likely in the TPA someplace), and will then attempt
       to parse whatever is left in TBUFF.  Sigh...  This explains
       why so many CPMUG programs do a SETDMA(TBUFF) before
       exiting to CP/M.

       3. When expanding the arguments in the .SUB file, the SUBMIT
       command actually writes the command lines, one per record,
       in reverse order!!!  The CCP notices the $$$.SUB file, opens it,
       reads the high-water-mark record, deletes it from the FCB,
       and closes the $$$.SUB file.  In reading thru the CCP,
       I get the feeling that $$$.SUB files won't work if they
       are larger than one extent, but I haven't verified this.
       Anyone have experience with this?

All in all, the SUBMIT/XSUB system would best be served up for dinner
on Thanksgiving day.

Dave Fylstra