Mailing List Archive

Martin,

Though I have never heard of this particular bug, this is definitely a
possible side effect of module loading order. I would like to think that
mod_backhand doesn't munge any of the variables it is passed. I was pretty
careful when I wrote it to not do this.

If jserv is not so careful, it could feasibly *change* some of the variables
in the request_rec struct for its own internal use. I am *not* claiming that
it does this, as I have never examined the code. It just seems intuitive that
if when mod_jserv's hooks are called before mod_backhand's it crashed then
something must be happening in mod_jserv's hooks.

The odd things is that the byHostname function doesn't really use any of the
servers variables. Hmmm...

Another thing -- if you are running on Solaris, things don't work right with
BackhandFromSO unless mod_bakchand is compiled statically into Apache. It
doesn't appreciate loaded modules loading other modules for some reason I
cannot determine.


Martin Sillence wrote:
> I'm testing the backhand module with the jserv module in an attempt to
> get load balancing of both static and dynamic pages.
> When loading the backhand module after the jserv I find that the
> byHostname sev faults on trying to access the hostname in the
> serverstats structure.
> This seems to go away when loading the module before the jserv.
> Is this a know feature of modules or a bug?

--
Theo Schlossnagle
1024D/A8EBCF8F/13BD 8C08 6BE2 629A 527E 2DC2 72C2 AD05 A8EB CF8F
2047R/33131B65/71 F7 95 64 49 76 5D BA 3D 90 B9 9F BE 27 24 E7

"Theo E. Schlossnagle" wrote:
>
> Martin,
>
> Though I have never heard of this particular bug, this is definitely a
> possible side effect of module loading order. I would like to think that
> mod_backhand doesn't munge any of the variables it is passed. I was pretty
> careful when I wrote it to not do this.
>
> If jserv is not so careful, it could feasibly *change* some of the variables
> in the request_rec struct for its own internal use. I am *not* claiming that
> it does this, as I have never examined the code. It just seems intuitive that
> if when mod_jserv's hooks are called before mod_backhand's it crashed then
> something must be happening in mod_jserv's hooks.

I was wondering if it might be a feature of the global nature of the
serverstats structure?
I'm a little suspicious of a global extern in a loadable module....
The passed parameters are OK.
I've not delved deeply have you tried it with any other modules? Is it
normal to have externs in modules?

--
Regards,
M

--
Martin Sillence / Unix/Internet developer / PR Newswire Europe Ltd.
martin.sillence@prnewswire.co.uk / Tel: 01865 785065 / Fax: 01865 785100
--
For support call 01865 785055
---
EMail Disclaimer: http://www.oit.net/prne_disclaimer.txt

The serverstats structure isn't actually externed, mod_backhand.c owns that
one. It is just externed in other source files so that they can see it.
serverstats *should* function just like the apache scoreboard -- as I took
most of the code from that.

I should be fine to have externs in a shared object, but as far as I know
mod_backhand doesn't have any.

As for the seg fault in byHostname, the only places I could imagin it
happening are around the lines where the free() occurs. Do you have any idea
on what line it is dying?

Martin Sillence wrote:
> I was wondering if it might be a feature of the global nature of the
> serverstats structure?
> I'm a little suspicious of a global extern in a loadable module....
> The passed parameters are OK.
> I've not delved deeply have you tried it with any other modules? Is it
> normal to have externs in modules?


--
Theo Schlossnagle
1024D/A8EBCF8F/13BD 8C08 6BE2 629A 527E 2DC2 72C2 AD05 A8EB CF8F
2047R/33131B65/71 F7 95 64 49 76 5D BA 3D 90 B9 9F BE 27 24 E7