jas@proteon.com (John A. Shriver) (11/29/88)
Basics: gnuemacs 18.50, running on MicroVAX-II, Ultrix-32 V2.2. I was starting gnuemacs (emacs -e rmail) to read in my mail (0.7 MB of it!). The machine decided to crash while the mail was being gathered (there was a fatal I/O error to the swap partition, presumably as gumenacs was growing it's memory use). The mail was lost. It was not in /usr/spool/mail/jas, it was not in ~/mail/RMAIL (my rmail-file-name), it was not in ~/mail/.newmail (the rmail-mode default-directory is ~/mail/). The only place it could be found was in the vmcore from the system core dump, and not in very useful form (hop skip and jump every 512 bytes). There seems to be some level of defensive programming missing. The original /usr/spool/mail/jas should not have gone away until gmuemace was highly confident that the copy thereof (.newmail ?) was soundly on disk. (Not just somewhere in the write cache of the filesystem.) The rmail gathering functions have proven to be very sound in the face of disks filling up, which happens regularly here.
kayvan@APPLE.COM (Kayvan Sylvan) (11/30/88)
Try looking in /lost+found for a file owned by you. This has happened to me once. The reclaimed file was reconnected to a file whose name was some strange number. ---Kayvan
jas@proteon.com (John A. Shriver) (11/30/88)
I looked there too. No dice. There really was no file open enough to contain the mail sucessfully. I would have thought that /usr/spool/mail/jas would not get cleared out until .newmail was sucessfully created (and flushed to disk if the system provides a call to do that). I'm not a good enough lisp reader to understand how the gather function really works.