[9fans] /sys/src/cmd/fossil/cache.c:/^abort

[9fans] /sys/src/cmd/fossil/cache.c:/^abort

[Lucio De Re]

This need not be the most recent version, late March...

The "abort();" seems like a bug catching command, it occurred while I
was creating a Venti snapshot (following the wiki instructions).  The
error message that preceeded it was:

	cacheLocalData: addr=0 type got 0 exp 0: tag got 4c0fe2a2 exp 22f90fe3

this on a pretty underpowered Pentium 100MHz (/dev/cputype == P54C 101)
with 32Meg of RAM, but a fair amount of swap:

	1975/4600 memory 4492/19600 swap

My first attempt at getting fossil working, is there anything else I
ought to supply?  I can leave the machine untouched, if so desired,
for as long as it takes to debug.

++L

Re: [9fans] /sys/src/cmd/fossil/cache.c:/^abort

[rsc]

Upgrade your fossil to the latest sources and binaries
and then try again.  I'm not too interested in tracking down
whether this is a fixed bug or not.  It looks like the block
got freed and reused prematurely.  It's likely to be a
fixed bug.

Russ

Re: [9fans] /sys/src/cmd/fossil/cache.c:/^abort

[Lucio De Re]

On Fri, May 02, 2003 at 11:21:55AM -0400, rsc@plan9.bell-labs.com wrote:
>
> Upgrade your fossil to the latest sources and binaries
> and then try again.  I'm not too interested in tracking down
> whether this is a fixed bug or not.  It looks like the block
> got freed and reused prematurely.  It's likely to be a
> fixed bug.
>
Thinking that my message had been lost, I tried again, and the problem
did not recur.  I did, however, restart Venti and Fossil both,
reformat the arenas and indeces, etc.  In particular, I had a vacfs
running, which could, I imagine, caused the inconsistency.  After the
restart, it wasn't there to interfere.

Your point about staying up to date is also valid.  It's a little hard
for me, but that's my problem.

++L

Re: [9fans] /sys/src/cmd/fossil/cache.c:/^abort

[Russ Cox]

Vacfs is read-only.  It wouldn't have caused any problems.