On Friday 18 February 2005 19:58, Jason Stubbs wrote: > What this means? If you have an eclass in your overlay that results in > different *DEPEND, KEYWORDS or whatever *and* you either use CVS as your > PORTDIR or your PORTDIR's set of eclasses' mtimes match those of CVS, > you'll be committing broken packages. > > There's no quick fix for the caching side of things. The easiest option > is to prevent eclasses from overlays being used at all. That would still > leave a hole - when an eclass differs but has the same mtime with regards > to CVS and portdir - but is the easiest and safest.
Correction: The eclass caching mechanism looks at the repository that the ebuild was defined in. Therefore, the only people that are at risk are those that use their CVS repository as PORTDIR or in PORTDIR_OVERLAY. So, here's the deal... Umm, I was going to write what the deal is, but I'm still trying to figure it out. :( Anyway, the thing I/we (I'll be gone for the weekend) need to know right now is what people are willing to sacrifice and what they are not in order to get a speedy (even if only temporary) resolution. Regards, Jason Stubbs -- [email protected] mailing list
