Again, keywords dropped without notification to the stricken arch teams. I am of a mind to drop HPPA support altogether because this has happened once too often to my liking. Is it really that hard to file a bug report when you do that version bump?
The plan is to get ghc 6.10.4 working, everything else is less interesting. It's become a bit more interesting though, now that readline-6 is going stable. If you want to help, you could carefully build the ghc 6.8.2-r1 build. The ultimate goal is 6.10.4 though, which is still hard masked due to the changes we make from time to time. If you don't mind making many binaries, you can go ahead and make ebuilds for that too. We do post bug reports for ebuilds we believe in, not wasted work, or you'd get three times as many.
Is bug 206643 fixed with 6.10.4 ?
I'll post a keywording bug shortly about the new ghc 6.10.4 ebuild that just got unmasked. And yes, bug #206643 should not be present for this ebuild. That bug is a combination of 1) old ghc/ghc-bin logic, 2) some more liberal logic that we had intended only to keep in the overlay but it accidentally got committed (afaik).
ppc and ppc64 have the latest versions keyworded.
I do not think this is ever going to be resolved. Please drop HPPA keywording at your earliest convenience.
Just noticed that hppa keywords has been dropped completely. It's unfortunate to see you go as we've just unmasked ghc 6.12.3 in portage this week and are about to ask arch teams to build ghc-bins so we can KEYWORD ghc for more arches. We've already got binaries for amd64, ppc64 and x86. It'll solve many issues for us if all arches are keyworded equally, so we'll do our best to help. This ghc release seems easier to get working and has fewer dependencies, also the haskell language team has doubled in work force (we're two devs now!).
(In reply to comment #6) > Just noticed that hppa keywords has been dropped completely. Yes, it wasn't that much work, as opposed to guarding the whole haskell situation from the arch team perspective, which I gave up on long ago. > It's unfortunate to see you go as we've just unmasked ghc 6.12.3 in portage > this week and are about to ask arch teams to build ghc-bins so we can KEYWORD > ghc for more arches. We've already got binaries for amd64, ppc64 and x86. > It'll solve many issues for us if all arches are keyworded equally, so we'll do > our best to help. If you want to help arch teams, then never drop keywords silently again, let alone repeatedly. It used to work fine when ghc was being slowly introduced to the portage tree but it ran amok after about a year, when all kinds of QA issues cropped up and nothing was done about them. You still have ~80 open bug reports to attend to, it seems, and putting another version into the tree won't magically fix the bugs left open. Good luck in your endeavours. -JeR
> If you want to help arch teams, then never drop keywords silently again, let > alone repeatedly. bug #206643 seems to be related. As I see there is no single POV on keywording unavailable binaries. : It used to work fine when ghc was being slowly introduced to > the portage tree but it ran amok after about a year, when all kinds of QA > issues cropped up and nothing was done about them. You still have ~80 open bug > reports to attend to, it seems, and putting another version into the tree won't > magically fix the bugs left open. Good luck in your endeavours. -JeR Do you mean bugs in ghc on hppa specifically, or haskell in general? And I just wanted to note, the changelog message (and commit message, I had to dig it here by luck) is misleading and does not contain the bug mention. Example: {{{ + 01 Jul 2010; Jeroen Roovers <jer@gentoo.org> quickcheck-1.0.ebuild, + quickcheck-1.1.0.0.ebuild, quickcheck-1.2.0.0.ebuild: + Drop support for ghc and darcs. }}} Would be even nicer to see at least dropped keyword there (to see reason to drop them would be perfect!). I also don't see a reason to remove (working?) hppa binaries from SRC_URI in older ghc ebuilds. If they will go away from mirrors resurrecting hppa support will be way harder. In case of ppc64 sole ancient 6.4.2 binary helped a lot (yes, it was marked stable, but didn't work with modern GCCs). Is it the same motive as in https://bugs.gentoo.org/show_bug.cgi?id=324511#c2 ? From your comments it certainly does not look as such.
Added alpha binary. ia64 is being built ATM. TODO: - ppc - sparc - bsd
Added ia64 binaries.
Added x86-fbsd binaries (built on i686-gentoo-freebsd8 by Naohiro Aota)
Readding ppc and sparc (missing binaries for ghc-6.12.3)
== Request for building binaries for arches == ghc is a haskell compiler written in haskell (so it needs binary form of ghc to build itself). Haskell team does not have access to ppc/ppc64/sparc hardware yet, so we ask you to build binaries for us (and help to resolve this bug) What you need to bootstrap ghc: 1. working installed ghc binary not later, than two versions back from bootstrapable one, in other words you need ghc-6.8.2 of ghc-6.10.4 to build this one. 2. add ~arch keyword to the local tree 3. run bootstrap script to get the binary: http://dev.gentoo.org/~slyfox/build-ghc-bin.sh and move result to distfiles in the scheme below. OR ACCEPT_KEYWORDS="~arch **" USE="-binary doc ghcbootstrap ghcmakebinary" emerge --buildpkgonly =ghc-6.12.3 4. When script is done you should add binary to the ebuild like others there: ... arch_binaries="$arch_binaries ppc64? ( mirror://gentoo/ghc-bin-${PV}-ppc64.tbz2 )" ... and try to install ghc without bootstrap keyword ('USE=binary emerge ghc' will just install binary you've just built) 5. Test installed ghc a litle: 5.1. run script, which checks minimal functionality: http://dev.gentoo.org/~slyfox/test_ghc_hello.sh 5.2. try to emerge dev-haskell/cabal (an example of haskell library), dev-haskell/alex(an example of haskell binary) and haskell-updater 6. Push ghc binary to mirrors space and commit an ebuild to the tree. 6.1 keyword working haskell-updater in the tree. I can be pinged on IRC for help to resolve various haskell runtime/build-time issues. Thanks! P.S.: == Shameless ask == In order to offload burden of building binaries for next ghc versions and keywording most packages symmetrically for all supported arches I (slyfox) am asking: * ppc/ppc64 and * sparc teams for access to boxes on those platforms in order to: * build ghc binaries solely by haskell team forces * running/fixing ghc testsuite (not in tree yet) on those platforms * keywording/debugging haskell related stuff there I promise not to drain the box CPU power and run all tasks under 'nice -n 19', as ghc chews up to 10 hours of building time. I tried to reach tgall@ by IRC/e-mail and ranger@ by e-mail, but didn't get any response. Should I take it as implicit refuse? Thanks again!
Marked ~ppc, should be good to go. Please let us know if you want a newer ghc stabilized.
Marked ~sparc (last arch in list). Compiler is able to: rebuild itself, build working darcs, build working cabal-install.
