On Saturday 22 July 2006 15:47, Anton Golubev wrote:
> Hello Nick,
>
> > No, you just cut too much before posting.
>
> Actually, I'm not. It was the full working minimal httpd.conf, which I
> cleaned up from unrelated options in order to proof myself with exact
> problem source. Other values are optional defaults.
Hmmm. It looked less than complete (I had to hack it just to load
the relevant modules - then had to add a virtualhost to get a problem).
> > What happens when you apply the (second) patch I posted?
>
> It was a crash in apr_hashfunc_default() with char_key-0x0.
> You probably haven't seen the backtrace just few lines below the
> sentence you cited. It definitely came to the list, since I received a
> my post back from it in the same form as I sent.
Aha! I missed that.
#3 0x08081d67 in ap_dbd_prepare (s=0x8daf5a0, query=0xffffffff <Address
0xffffffff out of bounds>, label=0xffffffff <Address 0xffffffff out of
bounds>)
at mod_dbd.c:150
That's what your previous traceback had: screwed up arguments to
ap_dbd_prepare. I really can't see how that comes from the code.
Are you on some unusual hardware or architecture? Did you compile
both both httpd and apr/apr-util on the machine it's running on?
#4 0x0808042d in authn_dbd_prepare (cmd=0xbfea7a10, cfg=0x8e06628,
query=0x8e06630 "sample query") at mod_authn_dbd.c:70
That (and below) looks fine.
> I'm a bit upset with this miscommunication. Should I send you traces
> privately?
Sorry, but a busy hack who has been staring at a screen too long may
miss things in your post!
Anyway, I don't see any reason for that, other than a build problem/
incompatibility (which could perhaps arise in the rather complex
APR_OPTIONAL macros).
Anyone else?
--
Nick Kew