[275] in testers
Re: undelete rtpc 6.4 afs
daemon@ATHENA.MIT.EDU (daemon@ATHENA.MIT.EDU)
Tue Nov 21 08:50:53 1989
From: geer@ATHENA.MIT.EDU
To: Jonathan I. Kamens <jik@PIT-MANAGER.MIT.EDU>
Cc: geer@ATHENA.MIT.EDU, testers@ATHENA.MIT.EDU
In-Reply-To: Your message of Mon, 20 Nov 89 19:29:29 -0500.
Date: Tue, 21 Nov 89 08:50:18 EST
> 1. Were you working in AFS, NFS, or local disk?
e40-342f-2% pushd ~/open/doc ; pwd
/afs/athena.mit.edu/user/g/geer/open/doc
> 2. Do you have undelete aliased to anything (i.e. are you passing it
> any arguments that I didn't see from your example)?
e40-342f-2% where undelete
/bin/athena/undelete
e40-342f-2% alias undelete
e40-342f-2%
> 3. Was there an undeleted version of Makefile in the directory? If
> so, what kind of file was it (regular file, sym-link, directory)?
yes, one form existent, another delete'd. i wished
to throw away the existent one by undelete'ng the
delete'd one.
> 4. Can you duplicate this? If so, can you tell me how to duplicate
> it?
no.
> If you find that you cannot duplicate this occurrence, my first guess
> about the cause would be that I am failing to check a malloc
> somewhere, resulting in a pointer to a bad string, resulting in an
> attempt to unlink "", resulting in the removal of the entire contents
> of the directory, as you have described.
i am constantly out of memory, multiple times per hour.
for example, cut and paste from xterm's frequently fails
(silently - grr) for want of memory; my console window
pops up all the time with xalloc failure; if i stay logged
in at athena, i cannot telnet to the host from home as there
is insufficient memory to get logged in.
your hypothesis, ergo, is a good fit for the observed data.
> Of course, it is difficult to find a bug that only manifests itself
> when the program is running low on memory, but I will put some effort
> into seeing if that is the cause.
i certainly understand that. my irritation is mostly because
my homedir is afs, but my homedir is too big for the afs backup
system so i am running sans protection...
--dan
