[Gnu-arch-users] [BUG] corruption after cacherev (archive checksum mismatch)

[Gavin Baker]

Greetings,

I seem to have come across a fatal bug in tla.

I committed some changes to my work and did a successful update on my
laptop.  At this point everything was fine.  I then did a cacherev to
make sure I had a cached pristine so I could do diffs while
unconnected.  It failed for some odd reason, so I just ran the same
command again.  I then got an error message about a mismatched archive
checksum.  Now I am very scared (this is my thesis!).  I have not
touched the working tree or the archive since the bug triggered (in case
we need to do some 'post-mortem' investigation).

This is running the following build of tla on Mac OS X 10.3.4,
connecting to a Solaris 8 server NFS volume via sftp:

lord%emf.net--2004/dists--devo--1.0--patch-9(configs/emf.net-tla/devo.tla-1.2)

./src   lord%emf.net--2004/package-framework--devo--1.0--patch-3
./src/docs-hackerlab   
lord%emf.net--2004/docs-hackerlab--devo--1.0--base-0
./src/docs-tla  lord%emf.net--2004/docs-tla--devo--1.0--patch-5
./src/hackerlab lord%emf.net--2004/hackerlab--devo--1.0--patch-37
./src/tla       lord%emf.net--2004/tla--devo--1.2--patch-115


I have actually seen this happen once before, going from a Linux box via
sftp to another Linux box, under Debian with latest tla as of May. 
Unfortunately I didn't get the transcript of the error occuring, so I
couldn't report a decent bug.  It remains unresolved (though I made damn
sure we had plenty of backups).


So here is the capture of the session showing the error.  I have changed
my email but the important details are left in:


scorpio:~/work/thesis gavinb$ tla changes 
* looking for gavinb%somewhere.net--2004-cs/thesis--dev--0.1--patch-4 to
compare with
* build pristine tree for
gavinb%somewhere.net--2004-cs/thesis--dev--0.1--patch-4
Enter passphrase for key '/Users/gavinb/.ssh/id_rsa': 
* from archive cached:
gavinb%somewhere.net--2004-cs/thesis--dev--0.1--patch-4
* comparing to gavinb%somewhere.net--2004-cs/thesis--dev--0.1--patch-4
A/ notes
A/ notes/.arch-ids
A  notes/.arch-ids/=id
A  notes/.arch-ids/notes.txt.id
A  notes/.arch-ids/questions_for_sol_9jun04.txt.id
A  notes/notes.txt
A  notes/questions_for_sol_9jun04.txt
scorpio:~/work/thesis gavinb$ tla cacherev
Enter passphrase for key '/Users/gavinb/.ssh/id_rsa': 
* from archive cached:
gavinb%somewhere.net--2004-cs/thesis--dev--0.1--patch-4
sftp_decode_status: Packet too short(1): possibly garbage from server?
scorpio:~/work/thesis gavinb$ tla cacherev
Enter passphrase for key '/Users/gavinb/.ssh/id_rsa': 
* from archive cached:
gavinb%somewhere.net--2004-cs/thesis--dev--0.1--patch-4

********************************
 MISMATCHED ARCHIVE CHECKSUM

  archive: gavinb%somewhere.net--2004-cs
  revision: thesis--dev--0.1--patch-4
  file: thesis--dev--0.1--patch-4.tar.gz

  expected: 6fdd7eb6da07f629ea9ec655c01831da
  got: d496052c4f77768aac27788e8f9db7d6

********************************


So it seems that something went wrong with the ssh connection (I can't
imagine what; everything else seemed to be working fine) but this caused
it to corrupt the local copy.  I can only assume (pray!) that the
archive itself is still intact.

Any more detail, dumps, whatever, you need to track this down - let me
know.

Thanks,

  :: Gavin

-- 
Gavin Baker // gavinb*antonym_org // Linux|Python|Esperanto|MIDI|Cheese