Re: [GNUnet-developers] Re: [patch] BDB crash

[Tom Barnes-Lawrence]

On Mon, May 12, 2003 at 01:31:34AM +0100, I wrote:
>   Done. BTW, I take it the other patch you just sent to the group
> at large was for clearing up the first bug I reported, and you
> hadn't very very suddenly fixed this one!

  WHOOPS! Sorry, my bad. Using mutt with extra headers showing, I read
the wrong timestamps on your emails. Realise now that other patch I
referred to was from several hours later. Reading the patch, it
appeared to catch the funny cursor value.

  Figured it wasn't meant to be cumulative with the patches you'd
been asking me to try, so I cleared out my old gnunet0.5.3 tree,
unpacked it again, and freshly applied just the bdb0508.diff and
your latest bdb0511.diff. Remade, reinstalled.

 Ran it and ... sure enough, still a segfault, but not where
I was expecting:

Program received signal SIGSEGV, Segmentation fault.
[Switching to Thread 3076 (LWP 25787)]
0x4018af7e in vfprintf () from /lib/libc.so.6

#0  0x4018af7e in vfprintf () from /lib/libc.so.6
#1  0x4018de52 in vfprintf () from /lib/libc.so.6
#2  0x4018963f in vfprintf () from /lib/libc.so.6
#3  0x4001c3b0 in LOG (minLogLevel=2, 
    format=0x40378480 "BDB: Can't create cursor: %s (%i)\n") at
logging.c:184
#4  0x40377764 in lowForEachEntryInDatabase (handle=0x806f770, callback=0, 
    data=0x0) at low_bdb.c:258


 Now, if I'm not mistaken, what you've done is:

LOG(LOG_ERROR, "BDB: Can't create cursor: %s (%i)\n", ret,
       db_strerror(ret));


 Tried to fprint an int as a string and a string as an int! :)

When I switch the arguments around...


May 12 02:13:54 BDB: Can't create cursor: DB_RUNRECOVERY: Fatal error, run
database recovery (-30989)
May 12 02:13:54 BDB panic, that's the end
May 12 02:13:54 __BREAK__ at logging.c:210

 So you've caught that successfully now.

 But, trying gnunet-check -a -n -V only says:
You selected verification only, will not fix problems!
Checking Content Database

==> Done checking 51 entries in content database.
Checking indexed files
==> Done with 0 indexed files.

I couldn't find any other similar tool. Oh, I suppose it'd be a db3 utility
it's talking about.. OK if I just delete the DB and start again fresh,
or is there any more to be learnt from the thing?

Tomble