Skip Menu |

Preferred bug tracker

Please visit the preferred bug tracker to report your issue.

This queue is for tickets about the CPAN-Reporter CPAN distribution.

Report information
The Basics
Id: 60663
Status: resolved
Priority: 0/
Queue: CPAN-Reporter

People
Owner: Nobody in particular
Requestors: mschwern [...] cpan.org
Cc:
AdminCc:

Bug Information
Severity: Wishlist
Broken in: 1.1800
Fixed in: (no value)



Subject: fork() before reporting tests
When installing a big wad of CPAN modules, reporting the test results slows things down. I figured it was the process of sending an email, but it seems even the new Metabase system has a delay. Not nearly as bad as it was with email, but this can get really bad on a dodgy network the result being that for big builds I'll turn it off. Since reporting a test has no effect on the install process it could be done in the background while CPAN.pm goes about its business. There's no need to inform the user of a successful report, but telling them about a problem becomes a... problem. How the subprocess communicates with the parents is a SMOP, but wedging the information into the CPAN interface isn't. Any problems could be reported as part of the end-of-command summary that normally reports any problems with installing modules. This would require the end of summary waiting for the children to finish which, in the worst case, is no worse than were it is now. Good idea?
Subject: Re: [rt.cpan.org #60663] fork() before reporting tests
Date: Mon, 23 Aug 2010 20:51:36 -0400
To: bug-CPAN-Reporter [...] rt.cpan.org
From: David Golden <dagolden [...] cpan.org>
On Mon, Aug 23, 2010 at 7:50 PM, Michael G Schwern via RT <bug-CPAN-Reporter@rt.cpan.org> wrote: Show quoted text
> There's no need to inform the user of a successful report, but telling > them about a problem becomes a... problem.  How the subprocess > communicates with the parents is a SMOP, but wedging the information > into the CPAN interface isn't.  Any problems could be reported as part > of the end-of-command summary that normally reports any problems with > installing modules.  This would require the end of summary waiting for > the children to finish which, in the worst case, is no worse than were > it is now.
The interesting thing about moving to HTTP over email is that we now have synchronous rather than asynchronous transport. One possibility will be to have the test client just serialize reports as files in a directory and let a separate process (e.g cron job) process them as network bandwidth allows. (Which is exactly what a local mail agent used to do for email. Sigh.) -- David