Front page | perl.perl5.porters |
Postings from January 2012
On Sun Apr 05 23:32:09 2009, Alain.Brossard@thomsonreuters.com wrote:
> This is a bug report for perl from alain.brossard@reuters.com,
> generated with the help of perlbug 1.36 running under perl 5.10.0.
>
> After about a week of running without problem our monitoring
> program dies and leaves as the only clue the following entry in
> its log file:
>
> Size magic not implemented.
>
> So I started it using the debugger:
> ==
> PERLDB_OPTS="NonStop frame=1 AutoTrace
> LineInfo=/Emt/Agents/gpg-monitor/logs/gpg-monitor-perl-trace.$$"
> export PERLDB_OPTS
> /usr/local/bin/perl -dS /Emt/Agents/gpg-monitor/bin/gpg-monitor.perl
> -f
> /Emt/Agents/gpg-monitor/config/config.xml -i PROD1 &
> ==
>
> The log file (huge) ended with:
> ==
> 218: lock @Results if($useThreads);
> 218: lock @Results if($useThreads);
> 219: while(my $event = shift @Results)
> 220: {
> entering threads::shared::tie::SHIFT
> ==
>
> Checking how often this code was used before it crashed:
>
> root%2-emtsuds005[103] grep 'entering threads::shared::tie::SHIFT'
> gpg-monitor-perl-trace.2411 | wc -l
> 525768
>
> I've already found the hard way that modules "threads" and
> "WWW::Curl::Easy"
> do not mix in Solaris 8 : multiple core dumps per day in different
> areas
> of
> libcurl. This does work better in Solaris 10, no core dumps, and it
> survives
> for about a week before dying without leaving any messages.
>
> Is there anything I can do to track this further?
>
> The code is internal but maybe the following snippets can help?
> use strict;
>
> use Getopt::Std;
> use Data::Dumper;
> use IO::Socket;
> use IO::File;
> use Time::HiRes;
> use threads;
> use threads::shared;
>
> my @Results : shared;
> ...
> while (ContinueRun($sleeptime))
> {
> my ($start, $stop);
> $start = time if($useThreads);
> # Refresh variables
> &RefreshVars();
>
> # Run test clients
> foreach my $k (keys(%JOBS))
> {
> my $r1 = $JOBS{$k};
> my $r = $$r1;
>
> $r->{Count} += 1;
> if($r->{Count} == $r->{Frequency})
> {
> if( $useThreads )
> {
> my $t = new threads \&runTest, $r1;
> $t->detach;
> $r->{Count} = 0;
> }
> else
> {
> &runTest($r1);
> $r->{Count} = 0;
> }
> }
> }
>
> {
> lock @Results if($useThreads);
> while(my $event = shift @Results)
> {
> my($pid, $status, $displayname, $name, $sev, $msg) =
> split(/,/,
> $event);
> ...
> }
> if(0 && $ENV{OS} =~ /Win/i)
> {
> $count +=1;
> ...
> }
> }
> sub runTest
> {
> my ($or, $rqueue) = @_;
> my $r = $$or;
> my $tstres;
> if($r->{Type} == 1) {
> if( not $useCurlEasy or $ENV{OS} =~ /^Win/i ) {
> $tstres = &runHttpCmdLine($or);
> } else {
> $tstres = &runHttp($or);
> }
> } elsif($r->{Type} == 2) {
> $tstres = &runTcp($or);
> }
> {
> lock @Results if ($useThreads);
> push @Results, $tstres;
> }
> }
>
>
This ticket has not received a response in nearly three years -- but
reviewing it tonight, I can see why.
Apart from the fact that the problem occurred on an OS (Solaris) to
which I have no access and to ways of building Perl (threads) with which
I'm not familiar, we don't have a complete program to run to try to
simulate the problem. Some of the module imports (Getopt::Std;
Data::Dumper; IO::Socket; IO::File) play no role in the code fragments
that follow.
Poster: Have you had any better results since the original post? Can
you give us a more self-contained program to help in debugging?
Thank you very much.
Jim Keenan
---
via perlbug: queue: perl5 status: new
https://rt.perl.org:443/rt3/Ticket/Display.html?id=64484
Thread Previous