On Sun, Oct 14, 2018 at 03:22:43PM +0200, Helmut Grohne wrote: > On Sun, Oct 14, 2018 at 02:47:36PM +0200, Adam Borowski wrote: > > Thanks! Turns out that, rather than n32 being a single bit of flags, > > there's a 4-bit bitield there that encodes the variant. > > Ok, I'm attaching the big endian variants.
Awesome. And indeed, I had done it wrong. Fixed, thanks! > I guess you want to have a test suite anyway. ;) Good point. Oh if just I knew of where to find a bunch of sample ELF objects for all these archs to test elf-test on... :) (Not in 0.14 yet, though -- I think I'd add mips*r6* to the main tool first.) > > I pushed the change to github for now; please say if your testing requires > > a full Debian upload. > > Since I'm doing all of this in fresh chroots, using uploaded packages is > substantially easier. Hope the samples help with reducing the number of > round trips. In any case, I don't see a need to rush this. From my pov, > it's moving complexity from rebootstrap elsewhere (i.e. into arch-test). I recently had a lesson that postponing a release of a piece of software by several years just because I want to add some more goodies wasn't such a good idea. Let's try to err on the "spam releases" side instead. Meow! -- ⢀⣴⠾⠻⢶⣦⠀ ⣾⠁⢰⠒⠀⣿⡁ 10 people enter a bar: 1 who understands binary, ⢿⡄⠘⠷⠚⠋⠀ 1 who doesn't, D who prefer to write it as hex, ⠈⠳⣄⠀⠀⠀⠀ and 1 who narrowly avoided an off-by-one error.