On Tue, Jul 17, 2007 at 09:20:20PM -0500, Andy Lester wrote: > >Excellent. Could you tell me more about the crucial bug? > > We're up to two now, I think. Here's the first for sure: http:// > rt.perl.org/rt3/Ticket/Display.html?id=44019 > > We're returning a pointer to a struct on the stack.
Ah yes. I nice subtle 'may crash sometime later on' issue that is a pain to debug the other way around. > >Paul and I were exchanging messages about that tonight. I definitely > >want to scan all projects more often, but there's a build dependency > >tracking non-scalability issue. > > Could we do something to minimize the amount of time it takes? Could > we give you a subset of files to scan? It's not an issue of number of files, cpu time, or the like. It's a matter of builds breaking when: Projects move from cvs to svn, bitkeeper, git, monotone etc, Projects change the version of automake or autoconf they require, Projects move their repo, Projects start requiring an additional library, Projects require new enviroment - files, or variables to be set up, etc... Unfortunately, every project is different. I'm working on schemes to handle that universally - so the more time I spend fixing broken builds, the further off the universal solution is. So, I don't want to over-promise the build frequency just yet. > >Paul said he > >believes you're going to be as OSCON next week? I'll be there all > > Yeah, I'm coming in Saturday night, and then leaving Friday. I'm > coming in early specifically to hang out and talk to people. I will get in Sunday night, and be there until Friday afternoon. > And of course you have to come hear me talk about technical debt. Sounds good. I can relate some stories for you about the 'defect debt' some projects accumulate when they aren't using static analysis... > Probably we should get together some time before Wednesday, > especially if you want to get stuff for your talk about the Perl > help, although they might be pretty far out in my mind to remember. Any comments for my talk will just be anecdotes. The main content is already assembled, but you may tell me good things to drive changes in Scan, or suggest material for future talks. > The Perl 5 codebase is RADICALLY different from the Parrot one, thank > goodness. Heh. That could be taken in different ways ;-) -- -- David Maxwell Coverity Open Source Strategist http://scan.coverity.com/
