[16:22] <infinity> cjwatson: So... This armhf linker mess leaves me with another interesting mess to sort out.
[16:22] <infinity> cjwatson: In the short term, we technically have two linkers (both symlinks to the same one, but two canonical locations)
[16:22] <infinity> cjwatson: mklibs can't handle that situation AT ALL, and I don't want to make it.
[16:23] <infinity> cjwatson: So, I'm thinking a bit of hackery in d-i/Makefile around where there's already some fudging of the Hurd linker.
[16:23] <infinity> cjwatson: The real kicker is that what gets done by mklibs is non-deterministic (or, rather, hard to guess, because it'll pick the first PI from the first binary that claims to have one)
[16:24] <cjwatson> I'm OK with that kind of hackery for now
[16:24] <infinity> So, I guess I just test for both, and whichever one exists, link the other two it.  Or something.
[16:24] <infinity> s/two/to/
[16:24] <infinity> I'll experiment locally once I have glibc built in the archive.
[16:24] <infinity> But yeah.  Something like that will happen soon. :P
[16:25] <infinity> Just a heads-up, if you'd prefer to ignore it and not vomit.
[16:25] <cjwatson> mklibs is pretty awful, but nothing else does the job ...
[16:25] <cjwatson> happy to ignore it :-)
[16:25] <infinity> Teaching mklibs to deal with multiple PIs seems pointless, unless we ever intend to have multi-arch d-i initrds (oh god, say no).
[16:25] <infinity> So, yeah, the Makefile hack seems saner anyway.
[16:25] <infinity> FSVO "sane".
[16:26] <cjwatson> I see no point in multiarch for d-i :-)
[16:26] <infinity> Also, if you're in the mood to vomit, check out the last patch on the end of http://lucifer.0c3.net/~adconrad/eglibc.debdiff
[16:26] <infinity> This is the madness that distro consensus has driven me to.
[16:27] <cjwatson> Ew
[16:27] <cjwatson> Wait, aren't those strcmps backwards?
[16:27] <cjwatson> Shouldn't it be strcmp() == 0?
[16:27] <infinity> Nope.
[16:28] <cjwatson> Oh, double negatives
[16:28] <infinity> Yeah.
[16:28] <infinity> When using strcmp, get into POSIX shell mode.
[16:28] <infinity> Confuses me every damned time.
[16:28] <infinity> "Wait, this is the one time when C is shell?"
[16:29] <cjwatson> No, it's not that that confused me, it's that the entire test is negated
[16:29] <cjwatson> (effectively, by way of continue)
[16:29] <infinity> Oh, that too, yeah.
[16:29] <infinity> It's all confusing. :P
[16:29] <infinity> But anything in C returning 0 for "a-ok" just fucks with me.
[16:29] <cjwatson> Er.  Like every system call?
[16:30] <infinity> Yes, anything. :P
[16:30] <infinity> I'm easily confused.
[16:30]  * cjwatson passes infinity libdbus
[16:31]  * infinity sidesteps.
[16:33]  * cjwatson considers fixing usb-creator for syslinux-legacy to get it off ev's plate, since I don't see a branch for it
[16:34] <cjwatson> need to move some code around a bit, though
[16:38] <infinity> Say, you speak fluent dpkg, right?
[16:38] <infinity> How many beers would it take to convince you to waste a half hour fixing http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=588077 for me? ;)
[16:38] <ubot2> Debian bug 588077 in dpkg "/usr/bin/dpkg-divert: --add --rename should not rename a file currently owned by --package" [Important,Open]
[16:42] <cjwatson> Er.  Need to be critical-path on slightly fewer things before promising anything.
[16:43] <infinity> Sounds familiar.
[16:43] <cjwatson> Probably not horribly painful though.
[16:43] <infinity> It even has a patch!
[16:43] <infinity> ... if only it wasn't in Perl, and dpkg-divert is now in C.
[16:43] <infinity> *cough*
[16:44] <cjwatson> And dpkg-divert has a test suite now, which should help.
[16:44] <cjwatson> Ask me after our rls-p-tracking >=High queue's at zero, and you can pay me a beer at UDS.
[16:44] <cjwatson> (Which I'm hoping will be no later than Tuesday)
[23:16] <cjwatson> ev: considering http://paste.ubuntu.com/931825/ - still need to test, but comments welcome, including "hands off, I'm working on it already"
[23:16] <cjwatson> ev: (I'm off tomorrow)