> Taking Stephen (and Joachim's) advice to heart: anybody building > Darcs packages who wants to comment? Salvatore and Florian are our > Windows and MacOS X packagers. Anything to add, guys?
Surely the most important things are the two points that Eric underlined. About the versioning: I feel that the "make final release, then bump if something is wrong" is the better compromise, because: 1) It shouldn't happen that something is wrong when we are at that point; when it happens, sacrificing a natural number is not a big issue. 2) Building a (good) package is more than just launching a script. You need to perform the test suite (every time, even if last time went well: s**t happens); you need to reperform the tests that went wrong on Windows just because it keeps file locks between processes, and a lot of time our "rm -rf temp" fails; you need to upload the files to the darcs site, try to download them to see if something went wrong (i.e. permissions). I'm sure that with an infrastructure like the Debian one everything is done automatically, but on Windows actually it isn't. 3) Even a micro-minor change like bumping version numbers is a change. Something could go wrong. 4) I bet that we'd have more "pre-release" testers if we make a "final binary", more than an RC one. Salvatore _______________________________________________ darcs-users mailing list darcs-users@darcs.net http://lists.osuosl.org/mailman/listinfo/darcs-users