* * * * *

                               Spurious signals

Mark's [1] webserver Seminole [2] has been having problems running on tower.
The architecture of Seminole is multi-threaded, but the threading library
under RedHat [3] 5.2 (and Linux 2.0) is basically broken, so Mark is having
to simulate that by using fork(). And it's not working that well. It'll run
for a few days, then the process accepting new connections will die for no
apparent reason, leaving the main process hung since it's not the one
accepting connections and it isn't aware that its child process died.

Mark, being primarily an embedded systems programmer [4] isn't that well
versed in all the picky details of POSIX (and especially signals—granted it
is a mess) so I wrote some code to trap every possible signal [5], log it to
syslogd and then exit. Hopefully we'll be able to track down what might be
happening.

I also pull out the state of the registers but that is very dependant upon
the compiler options used and is a complete hack (that as far as I know, only
works under the Intel platform without the -fomit-frame-pointer option under
the Linux 2.0 kernel.

Did I mention it being a hack?

[1] http://www.conman.org/people/myg/
[2] http://gladesoft.com/products/seminole/
[3] http://www.redhat.com/
[4] http://www.embedded.com/
[5] gopher://gopher.conman.org/0Phlog:2003/03/31/BigSignalJuJu.txt

Email author at [email protected]

You are viewing proxied material from gopher.conman.org. The copyright of proxied material belongs to its original authors. Any comments or complaints in relation to proxied material should be directed to the original authors of the content concerned. Please see the disclaimer for more details.