For the stable bindist, I plan to write a special batch mode strict dependency checker, and I'd love to hear your comments on this idea, and possibly suggestions on how to improve upon it.
How it works: it first removes all non-essential packages from the system; then it "fink install"s the package to be tested. The result of this (success/build failure/etc.) is saved in a file. Then we start over, i.e. all packages are removed again, and the next package is processed. This way we are guranteed to catch missing dependencies. We don't catch all possible mistakes of course, though. But at least more than in the normal way of building stuff. To elaborate a bit: this app will walk the package tree, and starts at packages that have no dependencies (besides the essential packages); once all of these are tested, we can proceed to test packages depending on these already built packages; and so on. We will slowly "climb up" the tree. The app stores all the information it gathers in a flat file, and it will be possible to resume testing later based on that file (it'll skip any package listed in there). Of course some stuff can't be tested easily with this. For example anything requiring user interaction, however, that is a bad thing for Fink packages normally. QT requires it (to confirm the license; is this really necessary?). And Atlas requires it too (makes sense in this case, though). Also, if a package that other packages depend on fails to build, we have a problem. E.g. assume X11 fails to build, then no pkg using X11 can be tested. In this case, user intervention is needed. Finally, I am not sure yet how to handle alternate dependencies, e.g. (emacs|xemacs) - maybe I'll just use the first, or whatever. Cheers, Max -- ----------------------------------------------- Max Horn Software Developer email: <mailto:[EMAIL PROTECTED]> phone: (+49) 6151-494890 _______________________________________________ Fink-devel mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/fink-devel
