[EMAIL PROTECTED] (Adam P. Harris) writes: > In article <[EMAIL PROTECTED]>, John Goerzen <[EMAIL PROTECTED]> writes: > > Martin Mitchell <[EMAIL PROTECTED]> writes: > >> John Goerzen <[EMAIL PROTECTED]> writes: > >> > around compiling all the i386 stuff for the other archs. But > >> nobody > goes around compiling the stuff from the other archs for > >> i386! So if > I suddenly do all my package development on Alpha, > >> the Alpha will have > the current versions, and perhaps the Sparc > >> and m68k too, but i386 > will be obsolete! Fix anybody? > >> > >> I have been compiling the enscript package for i386, which Hartmut > >> Koptein maintains on powerpc. So there are people doing this, it > >> just isn't widespread at the moment. > > > Excellent. Is there some sort of automated mechanism like the other > > platforms have? (That is, packages get automatically build on these > > other platforms)? > > I believe people use quinn-diff plus dbuild plus god knows ?
[Incidentally, m68k uses debbuild for humans and the build daemons use something even c00ler.] Quinn diff can't trivially handle i386, mostly because of binary-only NMUs, last time I checked. But since the only actual case of this phenomenon is still Hartmut (and Martin is, apparently, making his packages a non-issue), I haven't bothered to think about it very hard. If it really is a problem nowadays, I could, of course... [Off the top of my head: enforcing a new numbering policy for bin-only NMU's (e.g. 3.5-1 -> 3.5-1.0.1 (and not 3.5-1.1)) would solve the problem and would also solve the problem of bin-only NMU's being clobbered by source NMU's; I did mean to propose this to debian-policy several months ago, but apparently I never got round to it] -- James
