Created attachment 93631 [details]
sh -x ppp-up, with "strace -f" on ppp-watch

sh -x /etc/sysconfig/network-scripts/ifup-ppp
../networking/profiles/default/ifcfg-Olympus >foo 2>&1

with ifup-ppp modified to use strace on ppp-watch:
      exec /usr/bin/strace -f /sbin/ppp-watch "${DEVICE}" "$2"

/dev/input/ttyACM0 at 115200 .... are you sure this is correct?

[pid 12278] read(0, "ATZ\r\r\nOK\r\n", 1022) = 10
[pid 12278] select(1, [0], [], [], {0, 100000}) = 0 (Timeout)
[pid 12278] write(3, "ATZ", 3)          = 3
[pid 12278] write(3, "\n", 1)           = 1
[pid 12278] write(3, "OK", 2)           = 2
[pid 12278] write(3, "\n", 1)           = 1
[pid 12278] time([1060817690])          = 1060817690
[pid 12278] rt_sigaction(SIGPIPE, {0x420db0d0, [], SA_RESTORER, 0x420276f8},
{SIG_IGN}, 8) = 0
[pid 12278] send(5, "<30>Aug 13 16:34:50 wvdial[12278"..., 39, 0) = 39
[pid 12278] rt_sigaction(SIGPIPE, {SIG_IGN}, NULL, 8) = 0
[pid 12278] time([1060817690])          = 1060817690
[pid 12278] rt_sigaction(SIGPIPE, {0x420db0d0, [], SA_RESTORER, 0x420276f8},
{SIG_IGN}, 8) = 0
[pid 12278] send(5, "<30>Aug 13 16:34:50 wvdial[12278"..., 38, 0) = 38
[pid 12278] rt_sigaction(SIGPIPE, {SIG_IGN}, NULL, 8) = 0
[pid 12278] close(0)                    = 0
[pid 12278] fcntl64(0, F_GETFL)         = -1 EBADF (Bad file descriptor)
[pid 12278] close(0)                    = -1 EBADF (Bad file descriptor)
[pid 12278] write(3, "--> ", 4)         = 4
[pid 12278] write(3, "stdin not read/write and $MODEM "..., 39) = 39
[pid 12278] write(3, "\n", 1)           = 1

hmmm, two times close(0) ??? very strange...

Which version of redhat-config-network?

ah 1.2.15...

Did you try 
# ifup ifcfg-ppp0
or whatever ifcfg-Olympus is mapped to?
You may also try to use redhat-control-network..

I really should read the bugreport completly :)

could you please attach /etc/wvdial.conf ?

Created attachment 93639 [details]
/etc/wvdial.conf

Password and Username have been protected.

Yes, /dev/input/ttyACM0 is correct.  This is my USB modem, it is there and
turned on [power from USB; the READY LED is on], kudzu at boot has found and
configured it, it is listed in /etc/sysconfig/hwconf, module 'acm' and other usb
modules are loaded.  Minicom works (after symlink so that /dev/ttyACM0 resolves
to /dev/input/ttyACM0 [bug in minicom]).  Baud set at 115200 following the usual
rule-of-thumb that cpu<->modem speed for a modem with compression should be
about 2.5 times the line speed (56K; 43K typical for an actual connection which
minicom has observed.)

looks good... hmm.. 

and your ifcfg-Olympus?

Created attachment 93640 [details]
ifcfg-Olympus

PAPNAME has been protected.

Hardlinked:
/etc/sysconfig/networking/devices/ifcfg-Olympus
/etc/sysconfig/networking/profiles/default/ifcfg-Olympus
/etc/sysconfig/network-scripts/ifcfg-ppp0

looks good also...

could you try:

# wvdial --remotename ppp0 Olympus

This lets wvdial start pppd without any initscripts, just to see, if it dials..

# wvdial --remotename ppp0 Olympus
(as root, from multi-user GUI mode directly after power-off boot) does dial. 

However, interrupting wvdial with <CTRL>C (during modem speed negotiation), and
then invoking RedHat > System Tools > Network Device Control; selecting the
inactive ppp0 and clicking Activate; again produces "Failed to activate ppp0
with error 8".  The modem Activity light flickers.  The symptom seems to be that
ppp-watch loses logical sync when trying to reset the modem with ATZ.  The modem
replies "ATZ\r\rOK\r" all at once, which apparently confuses ppp-watch.

Trying to Activate using Network Device Control directly after power-on reboot
also dials.  However, clicking on the Cancel button (during modem speed
negotiation) produces an abort with python traceback; I will attach the debug
info.  After recovery by killing ppp processes, trying to Activate again gives
"Failed to activate ppp0 with error 8".

Created attachment 93641 [details]
python dump+traceback from Network Device Control

Click on Cancel during modem speed negotiation after dialing from Activate. 
This produces an Abort dialog, python traceback, and offers to save debug info.
 ppp session continues; connection is established.

ok, python traceback fixed in CVS... thx for this!

I don't know if this is related but I found a problem in
/etc/ppp/peers/ppp0 where the profile name did not match the
"Nickname". There is a related thread on
http://www.linuxquestions.org/questions/showthread.php?s=&postid=753335#post753335

People are getting a similar connection error. I had this connection
error and modified the said file and all was well. It somehow kept my
old values in this file instead of updating it.

I am using Redhat 9 with an external serial modem.

I hope this helps.

James.

So what was the cause of this bug?
How do I get my system fixed?

Sorry, I should have mentioned.  I am having this problem on a RH8
system I built for a friend.  Just wondering if there is a specific
file I should update.  From the discussion in this thread its clear
the problem is solved, 'in CVS.'  But Its not clear how I am to solve
this problem on my system.

Thank you!

dnoyeb: Use Fedora Core 1, 2 or now 3 ... not the old unsupported RH8

Ouch.  I understand.  I built this box for a friend who got her first
computer, an old computer from the city.  RH8 requirements were leaner
than RH9 at the time.  I can only imagine Fedora's requirements.

I'm going to give wvdial a try.  Looks like that part is not where the
bug is!?