Front page | perl.perl5.porters |
Postings from July 2006
On Mon, Jul 17, 2006 at 04:39:44PM +1000, Adam Kennedy wrote: > >I'm in two minds about it. (Ignoring for now the question of who has the > >time, > >or alternatively who has the money). The downside of cleaning is the > >danger of > >introducing subtle new bugs. > > This one is likely to come down to QA to detect any such bugs early. > With enough testing, they should be able to be as preventable as possible. Call me paranoid but I don't stake my life on QA. I was within 24 hours of shipping a 5.8.4 that would have stuffed various suidperl installations (because they don't use suidperl in the way that it's supposed to be used as described by the core docs). The regular QA didn't catch this. We added peephole optimisations for reverse sort that got out into a stable release and no-one spotted the bug for several months. Change is dangerous. > However, it would take a lot of CPU. I'd be happy to see something like > all of Phalanx as a suitable test... I'd certainly like to see this. Although I hope that it generates few false positives. I guess that there will be some, but initial "tuning" will resolve them. Nicholas ClarkThread Previous | Thread Next