Mailing List Archive

oops that was meant for Rob.

the error log shows SIGSEGVs, but the gdb debugger refuses to
give any clue as to where the problem was.

I might not be around this weekend, so I can't promise to respond
promptly.

rob

> From: Rob Hartill <hartill@ooo.lanl.gov>
> Date: Fri, 7 Jul 95 17:57:52 MDT
>
> oops that was meant for Rob.
>
> the error log shows SIGSEGVs, but the gdb debugger refuses to
> give any clue as to where the problem was.
>
> If you have a chance to look at this again, running the server under
> gdb with the -X option might make the debugger a bit more helpful.
> (Also, recompiling with -g would help a great deal).

I had CFLAGS = -g (it just doesn't show in the final linking stage)
I threw another in when defining CC to make sure - same results. Also
tried gcc instead of HPs c89 compiler - no help.


gdb -X ->

gdb: unrecognized option `-X'
Use `gdb --help' for a complete list of options.

=-=-==

(gdb) where
#0 0x800aa340 in _end ()
#1 0x800aa310 in _end ()
Cannot access memory at address 0x7b00e368.



I'm out of here for the rest of the weekend... in search of some
moutain bike trails in Colorado. There's a place called "Apache"
somewhere on the way - I'll see if I can snap a good sign.

rob
--
http://nqcd.lanl.gov/~hartill/

From: Rob Hartill <hartill@ooo.lanl.gov>
Date: Fri, 7 Jul 95 17:42:17 MDT
Precedence: bulk
Reply-To: new-httpd@hyperreal.com


I just compiled 5.2 and plugged it in, but it's not working..

It's failing to respond, error messages end up in the window
I start the server from.

Bleah! I may be able to wangle access to an HP-UX system to try to
figure this out over the weekend...

Looks the same with 5.1. Works okay with apache 0.7, so it can't be
a config problem

And I'm not aware of any significant changes to http_main.c or
http_request.c between 0.4 and 0.5 which would account for something
like this. (The only thing that changed seriously in http_main.c
from, say, 0.3 through 0.5.1 was adding the locking stuff, which
should be compiled in on HP-UX anyway).

compile flags..

c89 -Wall -DHPUX -o httpd alloc.o http_main.o http_core.o http_config.o http_request.o http_log.o http_protocol.o util.o util_script.o modules.o mod_mime.o mod_access.o mod_alias.o mod_cgi.o mod_include.o mod_dir.o mod_auth.o mod_auth_dbm.o mod_negotiation.o mod_userdir.o mod_common_log.o -lndbm

rst

From: Rob Hartill <hartill@ooo.lanl.gov>
Date: Fri, 7 Jul 95 17:57:52 MDT

oops that was meant for Rob.

the error log shows SIGSEGVs, but the gdb debugger refuses to
give any clue as to where the problem was.

If you have a chance to look at this again, running the server under
gdb with the -X option might make the debugger a bit more helpful.
(Also, recompiling with -g would help a great deal).

rst

From: Rob Hartill <hartill@ooo.lanl.gov>
Date: Sat, 8 Jul 95 9:04:47 MDT

I had CFLAGS = -g (it just doesn't show in the final linking stage)
I threw another in when defining CC to make sure - same results. Also
tried gcc instead of HPs c89 compiler - no help.

Hmmm... on some systems, -g is required as an option at the link stage
as well, or the debugging information is thrown away.

gdb -X ->

gdb: unrecognized option `-X'
Use `gdb --help' for a complete list of options.

Not quite... -X is an option to the *server*, which lets you do things
like:

% gdb httpd
GDB is free software and you are welcome to distribute copies of it
under certain conditions; type "show copying" to see the conditions.
There is absolutely no warranty for GDB; type "show warranty" for details.
GDB 4.13 (sparc-sun-sunos4.1.3_U1),
Copyright 1994 Free Software Foundation, Inc...
(gdb) run -X -d /server/root/directory
...

to get a *single* "server child" process running under the debugger,
doing the child_main accept loop. (Basically, it skips all the forks
in http_main, and takes the "child process" code path of each until it
finally winds up in the accept loop).

This is a lousy way to run a server (single threaded!), but it's a
good way to debug one, since you can connect with the client of your
choice (as opposed to, say, running the thing in inetd mode and typing
at it), and you get full access to the debugger command set.

Needless to say, the -X option should *only* be used when running the
server under the debugger; what you get otherwise is a machine that
can only handle one request at a time, a la MacHTTP, and which dies
when you try to restart it.

rst