Re: GHC 8.4.3
Congrats on this grand Haskell update! Ludo’.
Re: GHC 8.4.3
宋文武 writes: > Ricardo Wurmus writes: > >> Ricardo Wurmus writes: >> >>> Ricardo Wurmus writes: >>> Timothy Sample writes: > • There are 21 failing packages. They are agda, beast, corrode, > git-annex, ghc-aws, ghc-gnuplot, ghc-haddock-test, > ghc-hashable-bootstrap, ghc-indents, ghc-monadplus, > ghc-nats-bootstrap, ghc-packedstring, ghc-regex, ghc-regex-tdfa-rc, > ghc-statistics, ghc-vector-builder, ghc-wave, ghc-xmonad-contrib, > raincat, rapicorn, and shellcheck. Raincat is now fixed on wip-haskell. >>> >>> git-annex is also fixed now. >> >> agda is also fixed now. >> >> ghc-regex has had its last release in 2017, so this is an upstream >> problem, I think. ghc-regex-tdfa-rc probably fails just because >> ghc-regex cannot be built. >> >> ghc-monadplus was only needed for agda. I found that version 1.3 builds >> fine, but 1.4.x does not. >> >> ghc-nats-bootstrap builds fine. >> >> ghc-hashable-bootstrap builds fine. >> >> ghc-indents now builds fine (after disabling the tests). >> >> ghc-haddock-test is deprecated and should be removed. >> ghc-packedstring is deprecated and should also be removed. >> >> This leaves ghc-aws, ghc-gnuplot, ghc-statistics, ghc-vector-builder, >> ghc-wave, ghc-xmonad-contrib, and shellcheck. >> >> Any takers? > > Fixed 5, leaving ghc-aws and ghc-statistics (I got test fails for > ghc-happy). Thank you! I just fixed ghc-aws and ghc-statistics. I cannot reproduce the test failure for ghc-happy. I think we’re good to merge this. Any objections? If there are none I’ll merge this branch into master tomorrow. Thank you Tim and 宋文武! -- Ricardo
Re: GHC 8.4.3
Hi, Ricardo Wurmus writes: > ghc-regex has had its last release in 2017, so this is an upstream > problem, I think. ghc-regex-tdfa-rc probably fails just because > ghc-regex cannot be built. The impression that I got is that ghc-regex-tdfa-rc was a prerelease package for ghc-regex-tdfa, and that the Haskell community has come to prefer ghc-regex-tdfa over ghc-regex. > This leaves ghc-aws, ghc-gnuplot, ghc-statistics, ghc-vector-builder, > ghc-wave, ghc-xmonad-contrib, and shellcheck. Wow! Nice work. -- Tim
Re: GHC 8.4.3
Ricardo Wurmus writes: > Ricardo Wurmus writes: > >> Ricardo Wurmus writes: >> >>> Timothy Sample writes: >>> • There are 21 failing packages. They are agda, beast, corrode, git-annex, ghc-aws, ghc-gnuplot, ghc-haddock-test, ghc-hashable-bootstrap, ghc-indents, ghc-monadplus, ghc-nats-bootstrap, ghc-packedstring, ghc-regex, ghc-regex-tdfa-rc, ghc-statistics, ghc-vector-builder, ghc-wave, ghc-xmonad-contrib, raincat, rapicorn, and shellcheck. >>> >>> Raincat is now fixed on wip-haskell. >> >> git-annex is also fixed now. > > agda is also fixed now. > > ghc-regex has had its last release in 2017, so this is an upstream > problem, I think. ghc-regex-tdfa-rc probably fails just because > ghc-regex cannot be built. > > ghc-monadplus was only needed for agda. I found that version 1.3 builds > fine, but 1.4.x does not. > > ghc-nats-bootstrap builds fine. > > ghc-hashable-bootstrap builds fine. > > ghc-indents now builds fine (after disabling the tests). > > ghc-haddock-test is deprecated and should be removed. > ghc-packedstring is deprecated and should also be removed. > > This leaves ghc-aws, ghc-gnuplot, ghc-statistics, ghc-vector-builder, > ghc-wave, ghc-xmonad-contrib, and shellcheck. > > Any takers? Fixed 5, leaving ghc-aws and ghc-statistics (I got test fails for ghc-happy).
Re: GHC 8.4.3
Ricardo Wurmus writes: > Ricardo Wurmus writes: > >> Timothy Sample writes: >> >>> • There are 21 failing packages. They are agda, beast, corrode, >>> git-annex, ghc-aws, ghc-gnuplot, ghc-haddock-test, >>> ghc-hashable-bootstrap, ghc-indents, ghc-monadplus, >>> ghc-nats-bootstrap, ghc-packedstring, ghc-regex, ghc-regex-tdfa-rc, >>> ghc-statistics, ghc-vector-builder, ghc-wave, ghc-xmonad-contrib, >>> raincat, rapicorn, and shellcheck. >> >> Raincat is now fixed on wip-haskell. > > git-annex is also fixed now. agda is also fixed now. ghc-regex has had its last release in 2017, so this is an upstream problem, I think. ghc-regex-tdfa-rc probably fails just because ghc-regex cannot be built. ghc-monadplus was only needed for agda. I found that version 1.3 builds fine, but 1.4.x does not. ghc-nats-bootstrap builds fine. ghc-hashable-bootstrap builds fine. ghc-indents now builds fine (after disabling the tests). ghc-haddock-test is deprecated and should be removed. ghc-packedstring is deprecated and should also be removed. This leaves ghc-aws, ghc-gnuplot, ghc-statistics, ghc-vector-builder, ghc-wave, ghc-xmonad-contrib, and shellcheck. Any takers? -- Ricardo
Re: GHC 8.4.3
Ricardo Wurmus writes: > Timothy Sample writes: > >> • There are 21 failing packages. They are agda, beast, corrode, >> git-annex, ghc-aws, ghc-gnuplot, ghc-haddock-test, >> ghc-hashable-bootstrap, ghc-indents, ghc-monadplus, >> ghc-nats-bootstrap, ghc-packedstring, ghc-regex, ghc-regex-tdfa-rc, >> ghc-statistics, ghc-vector-builder, ghc-wave, ghc-xmonad-contrib, >> raincat, rapicorn, and shellcheck. > > Raincat is now fixed on wip-haskell. git-annex is also fixed now. -- Ricardo
Re: GHC 8.4.3
Timothy Sample writes: > • There are 21 failing packages. They are agda, beast, corrode, > git-annex, ghc-aws, ghc-gnuplot, ghc-haddock-test, > ghc-hashable-bootstrap, ghc-indents, ghc-monadplus, > ghc-nats-bootstrap, ghc-packedstring, ghc-regex, ghc-regex-tdfa-rc, > ghc-statistics, ghc-vector-builder, ghc-wave, ghc-xmonad-contrib, > raincat, rapicorn, and shellcheck. Raincat is now fixed on wip-haskell. -- Ricardo
Re: GHC 8.4.3
I'm not a guix expert yet, so I looked at updating GHC and I got way out of my depth real fast. I just wanted to say thanks. > Some of you may recall that I started work on updating all of our > Haskell packages to their LTS 12 versions, and making sure they all > build with GHC 8.4.3. Is this lts-12.10? ghc-8.6.1 was release a couple weeks ago, but I assume it will be added after lts-13.0 comes out. > There are 21 failing packages. corrode[1] should probably be removed. It hasn't been touched in over a year and a half and has been surpassed by c2rust[2]. I'm also pretty sure nobody except the author ever really used it as it had a lot of bugs and couldn't convert much C code. [1] https://github.com/jameysharp/corrode [2] https://github.com/immunant/c2rust
Re: GHC 8.4.3
Hi Joe, Joe Hillenbrand writes: > I'm not a guix expert yet, so I looked at updating GHC and I got way > out of my depth real fast. > > I just wanted to say thanks. You are very welcome. >> Some of you may recall that I started work on updating all of our >> Haskell packages to their LTS 12 versions, and making sure they all >> build with GHC 8.4.3. > > Is this lts-12.10? Very good question! I meant to mention this, in fact. I froze everything at 12.8, so that I wouldn’t have to chase a moving target. Once everything is merged, we can keep track of the latest LTS 12 updates like we normally do with the importer. > ghc-8.6.1 was release a couple weeks ago, but I assume it will be added > after lts-13.0 comes out. Hopefully! We did skip GHC 8.2.2, so it’s hard to say. >> There are 21 failing packages. > > corrode[1] should probably be removed. It hasn't been touched in over a > year and a half and has been surpassed by c2rust[2]. I'm also pretty sure > nobody except the author ever really used it as it had a lot of bugs and > couldn't convert much C code. That’s good to know. Thanks for mentioning this. > [1] https://github.com/jameysharp/corrode > [2] https://github.com/immunant/c2rust -- Tim
Re: GHC 8.4.3
Hi Tim, > Some of you may recall that I started work on updating all of our > Haskell packages to their LTS 12 versions, and making sure they all > build with GHC 8.4.3. I think this work is ready to be included now (on > a WIP branch, at least). Thank you so much! > The commits exist at > <https://gitlab.com/samplet/guix/tree/wip-haskell> under the > “wip-haskell” branch. There are 258 commits, and they have been rebased > on master (although that was a few days ago, but it should be fine > because the Haskell packages have been relatively stable). I figure > that 258 commits is too many for guix-patches, so I’m hoping that > someone (Ricardo?) can take them from the repo linked above. There are > about 450 packages that need to be rebuilt. I have fetched the branch, rebased it onto current master, and pushed it to “wip-haskell” on Savannah. > Here are some notes that you might find interesting if you rely on our > Haskell infrastructure. > > • There are 21 failing packages. They are agda, beast, corrode, > git-annex, ghc-aws, ghc-gnuplot, ghc-haddock-test, > ghc-hashable-bootstrap, ghc-indents, ghc-monadplus, > ghc-nats-bootstrap, ghc-packedstring, ghc-regex, ghc-regex-tdfa-rc, > ghc-statistics, ghc-vector-builder, ghc-wave, ghc-xmonad-contrib, > raincat, rapicorn, and shellcheck. Some of these are abandoned > upstream, but most of them just need some extra attention. I will > look through them (especially git-annex), but I would like the > stability of the build farm as I do it. (Also, help wanted!) “beast” and “rapicorn” belong together. “beast” depends on “rapicorn”; but they don’t seem to have Haskell dependencies. In any case these two packages are outdated and would need to be updated. It’s fine if these two are broken for now. > • The new Haskell no longer supports the “--allow-newer” flag, so it > has to be worked around with phases and “substitute*”. However, I > have introduced a “#:cabal-revision” keyword that tells the build > system to patch in newer revisions of the package’s Cabal file from > Hackage. This means that we can use this mechanism whenever > upstream has updated the dependency constraints (which is the main > reason we made so much use of “--allow-newer” in the first place). Very good. Thank you. > • I removed “ghc-mtl” since it is now included in GHC. Okay. I’ll try to go through the patches in the coming days and push fixes to some of the failures if I can. -- Ricardo
GHC 8.4.3
Hi all, Some of you may recall that I started work on updating all of our Haskell packages to their LTS 12 versions, and making sure they all build with GHC 8.4.3. I think this work is ready to be included now (on a WIP branch, at least). The commits exist at <https://gitlab.com/samplet/guix/tree/wip-haskell> under the “wip-haskell” branch. There are 258 commits, and they have been rebased on master (although that was a few days ago, but it should be fine because the Haskell packages have been relatively stable). I figure that 258 commits is too many for guix-patches, so I’m hoping that someone (Ricardo?) can take them from the repo linked above. There are about 450 packages that need to be rebuilt. Here are some notes that you might find interesting if you rely on our Haskell infrastructure. • There are 21 failing packages. They are agda, beast, corrode, git-annex, ghc-aws, ghc-gnuplot, ghc-haddock-test, ghc-hashable-bootstrap, ghc-indents, ghc-monadplus, ghc-nats-bootstrap, ghc-packedstring, ghc-regex, ghc-regex-tdfa-rc, ghc-statistics, ghc-vector-builder, ghc-wave, ghc-xmonad-contrib, raincat, rapicorn, and shellcheck. Some of these are abandoned upstream, but most of them just need some extra attention. I will look through them (especially git-annex), but I would like the stability of the build farm as I do it. (Also, help wanted!) • The new Haskell no longer supports the “--allow-newer” flag, so it has to be worked around with phases and “substitute*”. However, I have introduced a “#:cabal-revision” keyword that tells the build system to patch in newer revisions of the package’s Cabal file from Hackage. This means that we can use this mechanism whenever upstream has updated the dependency constraints (which is the main reason we made so much use of “--allow-newer” in the first place). • I removed “ghc-mtl” since it is now included in GHC. If you review the patches (thanks!) you might want to know: • I went through all of them carefully and I sincerely hope the error and typo count is low! Also, most of them are machine generated, so those are probably fine. • There are a few more interesting commits that deserve special attention. There are (minor) changes to the Stackage importer, there’s the “#:cabal-revision” stuff, and there are some old/unneeded packages that I deleted. Let me know if there is anything I can do to make these changes easier to work with. -- Tim
