On 2010-03-30 at 17:32:11 [+0200], Niels Reedijk <niels.reedijk@xxxxxxxxx> wrote: > On 29 March 2010 23:36, Ingo Weinhold <ingo_weinhold@xxxxxx> wrote: > > > > On 2010-03-29 at 22:38:34 [+0200], PulkoMandy <pulkomandy@xxxxxxxxx> > > wrote: > >> > I wasn't thinking of other files, but e.g. a catkeys file known to be > >> > broken. Currently it can just be excluded in the Jamfile; when using > >> > globbing the file would need to be removed from the folder. > >> > >> I never saw a file becoming broken on it's own, so it would still be > >> possible to either revert to a previous version or resync from hta. > >> The worst thing that can happen is a catalog being out of date, in this > >> case you get an half-translated application. But it still works. > > > > I adjusted the DoCatalogs rule accordingly in r36002, but commented out > > the > > globbing part. Apparently the worst case is that linkcatkeys fails with a > > "wrong fingerprint" error, thus breaking the build. > > I don't know how feasible it is, but is there a way to design the > catalog files in such a way that they are much less susceptible to > these kinds of issues? That's what I wonder, too. I don't really understand the purpose of the fingerprints anyway. If older catalogs still work with newer applications (just with possibly untranslated strings), I don't see why linkcatkeys fails here. A warning would fully suffice IMHO. CU, Ingo