OK, a proper fix for the issue (at least it fixes Ben's example) is in CVS. The bug was indeed mine (or maybe it once wasn't a bug and I then changed the definition of _relatives w/o proper checking for consequences, thus turning it into a bug - I dunno anymore :-) My responsibility anyway, for what it matters).
Essentially, $to_be_rebuilt should be reset to 0 whenever a package is rebuilt. In the case of a pkg with relatives (splitoffs+master), that means also the relatives should be marked as already built.
The bug was that in the case with relatives, while all relatives were marked as already-built, the pkg which just was rebuilt was *not* marked as rebuilt... which then could in some cases cause a double rebuild. Phew :-)
I also checked in some (I hope) improvements on two comments related to this, hopefully making them a little bit clearer.
Anyway, I noticed another problem now. I just wanted to test-build swi-prolog, which depends on its splitoff "swi-prolog-lite". So I didAlas, this problem persists. I checked, it did not occur in 0.17.4, so it is definitely a regression in 0.18.0
fink build swi-prolog
Which lead to fink telling me that it needs swi-prolog-lite for that! Not good, not good at all. Wonder if this is a regression in 0.18.0 or an old issue, hm.
Boy, I look forward to the day when we have regression tests for the dependency engine. This one is IMO really *the* central part of any regression suite for fink, it'll save us lots and lots of trouble, pain and work :-)
Cheers,
Max
------------------------------------------------------- This SF.net email is sponsored by: IBM Linux Tutorials. Become an expert in LINUX or just sharpen your skills. Sign up for IBM's Free Linux Tutorials. Learn everything from the bash shell to sys admin. Click now! http://ads.osdn.com/?ad_id=1278&alloc_id=3371&op=click _______________________________________________ Fink-devel mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/fink-devel
