Mon 29 Dec 2003 03:56:05 PM UTC, original submission:
Forwarded from Debian bug #224334:
If I use revision libraries instead of prestine source, then my understanding
is that it is very important that no changes occur to the revision library.
If I modify a file in the revision library, it is automatically
detected, and the library is considered currupt.
If however, I delete a file from the revision library (eg. by accidents-
once did a "mv" from the library instead of a "cp"; file system
curruption, etc) tla will not realize that the archive is now currupt
for "changes" or "commit" operations.
This results in a patch that says this is a new file, when it should
really be a changed file. I am not sure how bad this really is, but it
makes me nervous (this is only a test archive though).
For the "get", operation tla will complained that the file I deleted doesn't
exist; while perhaps not user-friendly, this does at least point at the tla
must somehow know that the file should exist:
bam@scrooge:~/tmp/tree$ tla get bam@snoopy.apana.org.au--test/test--devo--0.1--patch-2
- from revision library: bam@snoopy.apana.org.au--test/test--devo--0.1--patch-1
Error calling `vu_lstat' for "/home/bam/revision/bam@snoopy.apana.org.au--test/test/test--devo/test--devo--0.1/test--devo--0.1--patch-1/./a" (No such file or directory)
PANIC: I/O error
tla really needs to be able to detect the error though for the
changes and commit operations too.
I think this is important; as for tla to be considered robust it
needs to be able to cope with common errors (IMHO).
I suspect a similar situation might occur for files added to the
library, not tested.
|