[01:06] <infinity> doko : TBH, I'm more curious about why it didn't fail on amd64, not why it did fail on the other 3 arches...
[01:07] <doko> heh, you finally did wake up? :P
[01:08] <infinity> Hey, it's a weekend, give me a break. :)
[01:09] <doko> it does build locally, anyway, it's not that urgent
[01:10] <infinity> Nothing's urgent on a Sunday.
[01:10] <doko> true :)
[01:10] <infinity> If there was any urgency at all today, I'd be fixing Xorg.
[01:10] <doko> it' 1am, so good night!
[01:10] <infinity> But, y'know.  Watch me not care until tomorrow.
[01:10] <infinity> G'night. :)
[03:34] <lamont> doko: configure:21121: gcj-wrapper  Test.java
[03:34] <lamont> ../dist/configure: line 21122: gcj-wrapper: command not found
[03:34] <lamont> configure:21124: $? = 127
[03:35] <lamont> Build-Depends: tcl8.4-dev, procps [!hurd-i386] , fastjar [!hppa !mips !mipsel !hurd-i386] , gij [!hppa !mips !mipsel !hurd-i386] , libgcj-dev [!hppa !mips !mipsel !hurd-i386] 
[03:35] <lamont> although that just explains the failure on hppa...
[03:36] <lamont> please remove the !hppa from all those doko.  kthxbye
[06:14] <fabbione> morning
[06:44] <fabbione> jbailey: interesting:
[06:44] <fabbione> Unpacking replacement libc6 ...
[06:44] <fabbione> FATAL: kernel too old
[06:44] <fabbione> but right :)
[03:09] <doko> infinity, lamont: please could you get the config.log file from the failed db4.2 builds?
[03:48] <infinity> doko : Yeah, I'll have to spin it again.  Gimme a few minutes.
[03:49] <infinity> doko : Yeah, I'll have to spin it again.  Gimme a few minutes.
[04:01] <infinity> doko : rookery:~adconrad/ has the failed tree.  config.log doesn't appear terribly enlightening, though.
[04:14] <doko> infinity: hmm, and is it possible to verify, that /usr/bin/java exists?
[04:14] <doko> or I make an upload and just add an ls
[04:15] <infinity> If it's shipped in the build-deps, it exists...
[04:15] <doko> gij-4.0 adds an alternative to java
[04:16] <infinity> Ahh, I wonder if the alternatives are broken, due to a broken package in the past.
[04:16] <infinity> That can happen.
[04:16] <infinity> Let me stop the buildd and fiddle in the chroot.
[04:17] <infinity> And that's the issue, alright.
[04:17] <infinity> In a clean chroot:
[04:17] <infinity> root@vernadsky:~ # update-alternatives --display java
[04:17] <infinity> java - status is manual.
[04:17] <infinity>  link currently points to /usr/lib/jvm/java-gcj/bin/java
[04:17] <infinity> No versions available.
[04:17] <infinity> So, somewhere in the past, a broken package fucked the alternatives.
[04:18] <doko> looks like java-gcj-compat ...
[04:18] <infinity> You may want to see if the package in question is still buggered.
[04:18] <infinity> But I can hit all the chroots and manually reset the alternatives to automatic to fix it for now.
[04:18] <infinity> (But if a package is breaking it, that package may break it again when it's installed)
[04:19] <jbailey> fabbione: Mmm, yeah.  I forgot the preinst guard, thanks.
[04:19] <jbailey> doko: I've been thinking of adding something to cdbs where if the configure step fails that it spews out config.log to stdout (so it'll get caught by the build log)  Interested?
[04:19] <doko> yes, found it :-/ wasaaaaabi !!!!
[04:20] <infinity> doko : Alright, I'm resetting all the chroots' alternatives and will respin db4.2 for you.  If you want to fix gcj-compat so this doesn't happen again, that'd be nice. :)
[04:20] <doko> jbailey: yes, sounds nice, but it doesn't catch the alternatives thing
[04:20] <jbailey> doko: Shouldn't the list of packages installed at the top catch that?
[04:21] <doko> ?
[04:21] <jbailey> The top of the logs shows which packages were installed.
[04:21] <infinity> jbailey : If a package fails to remove its alternative, things can go seriously tits-up.  The alternatives system is a bit fragile.
[04:21] <infinity> jbailey : There are bugs in the Debian BTS about this.  It breaks buildds from time to time.
[04:22] <jbailey> Oh, I see.
[04:22] <jbailey> Oh, I see.
[04:22] <jbailey> bah
[04:22] <infinity> In fact...
[04:23] <infinity> doko : Didn't the amd64 build complete because gcj-compat was still installed (dirty chroot, I suspect)?
[04:23] <infinity> doko : I remember the log looked different for amd64 than other arches, WRT java stuff.
[04:26] <doko> infinity: I didn't look at the amd64 log ;)
[04:30] <infinity> There, all the chroots fixed.  Respinning the builds now.
[04:32] <infinity> Woo.
[04:33] <infinity> doko : Did you break gij between the last db4.2 builds and now? :)
[04:33] <infinity> gij: Depends: gij-4.0 (>= 4.0.0-7) but it is not going to be installed
[04:34] <doko> infinity: yes, fixed in -ubuntu2
[04:34] <doko> infinity: yes, fixed in -9ubuntu2
[04:34] <doko> I did not break it, I did improve it.
[04:35] <infinity> Ahh, that version is still building on powerpc and ia64.  Feh.
[04:36] <doko> yes, we should have a ppc64 kernel on the buildd's
[04:37] <infinity> We'll get there soon.
[04:37] <infinity> Need to get elmo to install said kernel, and we need to unilaterally linux32 buildd on all the buildds, on the off chance that packages do Very Bad Things when they see they're running on ppc64.
[04:38] <infinity> Not a lot of work, just needs some co-ordination this week, and someone to prioritise it.
[04:39] <infinity> Wow, creepy.  powerpc and ia64 are at exactly the same place in the testsuite right now.
[04:40] <infinity> (Which is "nowhere near the end")
[04:40] <fabbione> jbailey: no no.. the test is fine.. it detected correctly a 2.4 kernel
[04:41] <infinity> doko : db4.2 was happy on i386, though, where gcc-4.0 is up to date.
[04:42] <doko> cool, thanks
[04:43] <fabbione> oh crap
[04:43] <fabbione> doko: another gcc upload eh?
[04:44] <infinity> doko : Any other weird failures on your "to investigate" list?
[04:44] <doko> fabbione: no
[04:44] <doko> next one is 4.0.1 when it gets released
[04:45] <infinity> Yay!
[04:45] <fabbione> doko: you upload 9ubuntu2
[04:45] <infinity> What's the time frame for 4.0.1?
[04:45] <fabbione> don't lie!
[04:45] <fabbione> ;)
[04:45] <fabbione> i don't think we will manage to get it before UVF
[04:45] <fabbione> doko: UVF is in less than 2 weeks
[04:45] <doko> fabbione, you didn't ask for the version :P
[04:46] <doko> fabbione: we will. release should be next week
[04:46] <fabbione> doko: well after 12 hours in the build..gettting the other one is a pain
[04:46] <fabbione> given that all the build time is wasted
[04:46] <infinity> Aww, pumpkin.
[04:46] <infinity> FEEL MY PAIN.
[04:47] <fabbione> infinity: we need to find a way to bootstrap ghc6 :)
[04:47] <fabbione> right now we are in one big unbreakable loop
[04:47] <infinity> Yu and ghc are having issues?
[04:47] <infinity> I remember bootstrapping it not that long ago, actually.
[04:47] <infinity> On Debian, but same shit.
[04:47] <fabbione> no ghc6 is having issues everywhere
[04:47] <fabbione> infinity: wanna better in Ubuntu is more complicate?
[04:47] <fabbione> s/better/bet
[04:48] <infinity> Ian and I went through the pain of bootstrapping it on m68k together, IIRC.
[04:48] <fabbione> yes but you had ghc5 available.. didn't you?
[04:48] <infinity> If it mixes and matches stuff from main and universe, that could make it worse.  Otherwise, it should be exactly the same pain.
[04:49] <infinity> No, ghc5 only existed on i386.
[04:49] <fabbione> hm ok
[04:49] <infinity> All other ghc6 arches had to be bootstrapped by hand.
[04:49] <fabbione> because upstream bootstrap is broken as it is
[04:49] <fabbione> and of course nobody cared to fix it while bootstrapping
[04:49] <infinity> Yeah.  This is known.
[04:49] <infinity> I'd recommend getting Ian involved.
[04:50] <infinity> Unless he's an Ubuntu-hating nut (and I doubt it), he's pretty keen on making ghc6 widely-available and bootstrappable.
[04:50] <fabbione> does he do irc?
[04:50] <infinity> No idea...
[04:51] <fabbione> because we might have to bootstrapping it again on all arches
[04:51] <infinity> Yup, igloo.  Idle 11 mins.
[04:51] <infinity> Should be around.
[04:51] <fabbione> ok
[04:52] <infinity> I just invited him to come play in here.
[04:52] <fabbione> ah ok
[04:52] <\sh> could someone be so nice and check this out? http://people.ubuntu.com/~lamont/buildLogs/h/hdf5/1.6.4-2ubuntu1/hdf5_1.6.4-2ubuntu1_20050617-1417-powerpc-failed.gz
[04:54] <infinity> I make it a policy not to look at build logs with the word "libtool" in the failure...
[04:56] <fabbione> good policy
[04:56] <infinity> In the tradition of "all things powerpc suck until we get a new kernel", I'm just going to retry it for you.
[04:56] <fabbione> i use the same :)
[04:56] <fabbione> infinity: .12 final is out.. i hope i can get it in the archive tomorrow or tuesday
[04:57] <fabbione> after that i am pretty sure elmo will be quite happy to install it around
[04:57] <infinity> fabbione : Rock.
[04:57] <infinity> fabbione : My PPC mahcine gets here in a day and a half!  Yay!
[04:57] <fabbione> doko: do you plan to upload another gcc-4.0 in the next 24 hours?
[04:57] <doko> no
[04:58] <fabbione> doko: are you sure?
[04:58] <doko> no
[04:58] <fabbione> doko: are you ABSOLUTELY sure?????
[04:58] <doko> :)
[04:58] <doko> 4.0.1 won't get released tomorrow ...
[04:59] <fabbione> ok
[04:59] <Igloo> Hello
[04:59] <fabbione> hi Igloo 
[04:59] <infinity> fabbione : You're the man with the questions, Igloo's (hopefully) the man with the answers.
[04:59] <Igloo> If you're worrying about ghc6 6.2.2 not being able to buld 6.4, that's hopefully a rare event
[04:59] <fabbione> Igloo: we have some ghc6 fun atm.. i was hoping you could help :)
[04:59] <infinity> Igloo : More about scorched-earth bootstrapping.
[05:00] <fabbione> Igloo: we need to bootstrap from scratch :)
[05:00] <fabbione> because of libgmp3 c++ transition.. ghc6 becomes uninstallable
[05:01] <fabbione> so we need to rebootstrap ghc6 to build on top of the new libgmp3c2
[05:01] <fabbione> but my attempts are failing miserably.
[05:01] <fabbione> infinity: remeber you will have to do the same dance on i386/ppc/amd64/ia64 :)
[05:02] <infinity> A static compile of the old ghc would solve that problem.  But we still need to be able to do a scroched-earth bootstrap anyway, so it's worth investigating.
[05:02] <fabbione> Igloo: so i was wondering if you have time to go trough it with me
[05:02] <infinity> scorched, even.
[05:02] <Igloo> Ah, rats. Is libgmp entirely C++, or is it possible ghc is only using C bits?
[05:02] <fabbione> libgmp3-dev Depends: libgmp3c2
[05:02] <fabbione> i think it's pure c++
[05:03] <fabbione> Igloo: but i was more concerned to make the overall clean..
[05:03] <infinity> Not true.
[05:03] <doko> Igloo: perl -pi -e 's/libgmp3/libgmp3c2 | libgmp3/' debian/<pkg>.substvars
[05:03] <infinity> It's two libs in one package, cause the maintainer SUCKS.
[05:03] <fabbione> like being able to bootstrap, setting a var in debian/rules
[05:03] <infinity> It's a C lib and a CXX lib.
[05:03] <doko> fabbione: no
[05:03] <Igloo> Have you seen the instructions on http://www.haskell.org/ghc/docs/latest/html/building/sec-porting-ghc.html#unregisterised-porting ? Those should work if nothing else does
[05:04] <doko> gmp (4.1.4-6.1) unstable; urgency=high
[05:04] <doko>   * NMU coordinated with the maintainer. Upload as a build dependency
[05:04] <doko>     for building gfortran-4.0 with the changed library name.
[05:04] <doko>   * Split out the C++ library in it's own package libgmp3++. The libgmp3
[05:04] <doko>     package is renamed to libgmp3c2, or else packages relying on the C++
[05:04] <doko>     library in libgmp3 will break.
[05:04] <doko>   * Let a dependency on libgmp.so.3 be satisfied by 'libgmp3c2 | libgmp3',
[05:04] <doko>     allowing transition of packages to etch, which do not depend on the C++
[05:04] <fabbione> Igloo: no i did try to use several ./configure options... but i can look at it
[05:04] <doko>     library, before libgmp3c2 and libgmp3++ reach etch.
[05:04] <doko>   * Explicitely build using g++-4.0, build for i486, not i386.
[05:04] <doko>   * Call dh_makeshlibs in the install target, not the binary-common target.
[05:05] <infinity> doko : Ah-ha.  Did you do that?
[05:05] <infinity> doko : If so, I love you.
[05:05] <doko>  -- Matthias Klose <doko@debian.org>  Sun, 12 Jun 2005 10:32:34 +0200
[05:05] <infinity> doko : Why did you never upload? :)
[05:05] <fabbione> doko: see.. don't sell me shit you uploaded this morning in Debian :)
[05:05] <doko> waiting for binutils ...
[05:06] <infinity> Ah.
[05:06] <infinity> Well, can we get that uploaded to Ubuntu, like, two days ago?
[05:06] <fabbione> doko: because iam trying to work with what's in the archive :)
[05:06] <infinity> (Or has it already?)
[05:06] <doko> nah.
[05:07] <infinity> Did you split all three libs, or just the CXX lib?
[05:07] <Igloo> /usr/lib/ghc-6.4/ghc-6.4 is linked against libgmp.so.3 but not libgmpxx.so.3, so I'd guess it would work to just forcibly install the new libgmp3 and old ghc6 together to build the new ghc6?
[05:07] <doko> infinity: just C++
[05:07] <infinity> Igloo : Yes, that would work for the problem at hand.
[05:08] <infinity> Igloo : Of course, the question of scorched-earth bootstrapping is still a curious one. :)
[05:08] <fabbione> Igloo: yeah but we don't have the previous ghc6 on all arches.. yet :)
[05:08] <Igloo> fabbione: The above URL should help you with that. Or just bootstrapping from the Debian packages.
[05:09] <fabbione> Igloo: i am bootstrapping from the Debian packages.
[05:09] <Igloo> infinity: First you'll need to write a Haskell compiler in C which supports the extensions GHC's source uses.
[05:09] <infinity> Hrm, good point.  We're missing it on ia64 (and I assume sparc and hppa?)
[05:09] <Igloo> Oh...do you have more arches than Debian then?
[05:09] <fabbione> infinity: we can probably use ghc6 from Debian, but that means doing a lot of extra rebuilds
[05:10] <infinity> Igloo : No, fewer.  We have amd64, i386, and powerpc officially (and those all have ghc6 builds already), and ia64, sparc and hppa unofficially (none of which do, I suppose, but we have Debian packages for all of them)
[05:10] <infinity> fabbione : A two-pass bootstrap should be good enough.
[05:10] <fabbione> infinity: for 2/3 reiterations required to build ghc6 and haddock (and another 2 pkgs iirc)
[05:10] <fabbione> infinity: yes but it is extended to more than ghc6 :(
[05:10] <infinity> fabbione : But I'll test this theory on ia64, since I need to anyway.
[05:10] <fabbione> that's why i would prefer direct bootstrap
[05:11] <infinity> How many packages are stuck this way right now?
[05:11] <fabbione> Igloo: only i386/ppc/amd64 are official
[05:11] <fabbione> Igloo: the others are unofficial ports
[05:11] <infinity> doko : And can we get that split gmp upload done as 6ubuntu2 ASAP?
[05:11] <fabbione> infinity: i think it's a 3 pkgs loop. ghc6 haddock and a b-d for haddock (that b-d on ghc6
[05:12] <infinity> That doesn't sound so bad.
[05:12] <fabbione> infinity: so i think it's easier to just bootstrap ghc6 from scorched earth
[05:12] <fabbione> infinity: and break the overall
[05:13] <infinity> Where does haddock come in?
[05:13] <infinity> I don't see it as a build-dep for ghc...
[05:13] <Igloo> haddock is used to build ghc's documentation
[05:13] <doko> infinity: yes, do you need it now? not really. and in ubuntu, libgmp3c2 needs to depend on libgmp3++ anyway.
[05:13] <fabbione> infinity: it's in the b-d
[05:13] <infinity> Oh, docs are no big deal.  That's part of a 2-pass bootstrap.
[05:13] <infinity> Like we need arch:all stuff in a bootstrap compiler.
[05:14] <infinity> doko : Not if we do a mass-rebuild to fix the deps in Ubuntu.  We have this poower.
[05:14] <infinity> doko : Breaking breezy is not only okay, it's encouraged.  So we should do it.
[05:15] <infinity> doko : I see no reason (except for breezy->breezy upgrades, which are so NOT supported right now) for libgmp3c2 to depend on libgmp3++
[05:18] <infinity> Igloo : All this aside, is there any reason why the ghc6 packages can't currently do a pure bootstrap on their own the handholding of that readme you pointed us to? :)
[05:18] <infinity> s/their own/their own without/
[05:19] <fabbione> make[1] : *** No rule to make target `Apply.o', needed by `libHSrts.a'.  Stop.
[05:19] <fabbione> make: *** [all]  Error 1
[05:19] <fabbione> that's the same place where i crashed before
[05:19] <fabbione> Don't worry if the build falls over in the RTS, we don't need the RTS yet.
[05:19] <fabbione> AH
[05:19] <fabbione> crap
[05:20] <fabbione> yeah but later libraries don't build
[06:09] <infinity> \sh : hdf5 on powerpc went fine with a retry, by the way.
[07:47] <infinity> doko : Still around?
[07:49] <infinity> doko : That dpkg build-dep is utterly useless.  I assume you wanted a versioned build-dep on dpkg-dev, not dpkg...
[07:49] <infinity> doko : Not worth an upload to fix, since the Ubuntu chroots are all up to date anyway, but worth fixing in your local repo before you do a -10 upload to Debian, or another Ubuntu upload.
[07:50] <doko> infinity: ok, will fix it.
[07:50] <infinity> Danke.
[08:09] <fabbione> hmmm
[08:09] <fabbione> of what package?
[08:09] <fabbione> infinity: you will need gcc-3.3 for ghc6
[09:36] <doko> fabbione: did you have fun on the TV ;-)
[09:39] <lamont> historically, we've just bootstrapped from debian if that's an option
[10:48] <fabbione> doko: haven't seen the race yet.
[10:48] <fabbione> they didn't broadcast it yet
[10:48] <fabbione> (fsckuers)
[10:49] <fabbione> lamont: doesn
[10:49] <fabbione> lamont: doesn't build anyway :)
[11:09] <fabbione> bah
[11:09] <fabbione> they did broadcast only the first lap
[11:09] <fabbione> bastard
[11:17] <zul> fabbione: you are up late
[11:18] <fabbione> zul: i was hoping to see the F1 race
[11:19] <zul> ah..
[11:19] <zul> let me guess ferrari fanboy right?
[11:19] <fabbione> yeah
[11:21] <fabbione> and than they show the first lap and stop bradcasting
[11:22] <fabbione> danish tv is the sucks
[11:22] <fabbione> i need to get some satellite stuff
[11:24] <fabbione> i think i am going to get some sleep
[11:24] <fabbione> cya tomorrow
[11:31] <jbailey> fabbione: Right, but for all the other archs there's a preinst check that verifies the kernel version and gives a sensible error message.