Summary: | dev-haskell/haddock-2.14.3.0.7.8.3 fails install phase - setup: Error: Could not find module: Documentation.Haddock with any suffix: ["dyn_hi"] in the search path: ["dist/build"] | ||
---|---|---|---|
Product: | Gentoo Linux | Reporter: | Alexandre Rostovtsev (RETIRED) <tetromino> |
Component: | [OLD] Development | Assignee: | Gentoo's Haskell Language team <haskell> |
Status: | RESOLVED DUPLICATE | ||
Severity: | normal | ||
Priority: | Normal | ||
Version: | unspecified | ||
Hardware: | All | ||
OS: | Linux | ||
Whiteboard: | |||
Package list: | Runtime testing required: | --- | |
Attachments: |
emerge --info + haddock build.log
tree /var/tmp/portage/dev-haskell/haddock-2.14.3.0.7.8.3 verbose build log |
Description
Alexandre Rostovtsev (RETIRED)
2014-12-30 05:59:08 UTC
That looks very strange. Can you reproduce it consistently?
ghc obviously compiled a source file:
> [41 of 41] Compiling Documentation.Haddock ( src/Documentation/Haddock.hs, dist/build/Documentation/Haddock.o )
It should have been created Haddock.o Haddock.hi Haddock.dyn_o Haddock.dyn_hi.
Do you have it there?
Seems this issue related to upstream one https://ghc.haskell.org/trac/ghc/ticket/9176 and it is not solved yet. For me dyn_hi is not created for an 'empty' modules that have only reexports, like this one in haddock. Created attachment 392700 [details] tree /var/tmp/portage/dev-haskell/haddock-2.14.3.0.7.8.3 (In reply to Sergei Trofimovich from comment #1) > That looks very strange. Can you reproduce it consistently? Yes. Tried 4 times. > ghc obviously compiled a source file: > > [41 of 41] Compiling Documentation.Haddock ( src/Documentation/Haddock.hs, dist/build/Documentation/Haddock.o ) > > It should have been created Haddock.o Haddock.hi Haddock.dyn_o > Haddock.dyn_hi. > Do you have it there? Yes, I have Haddock.o Haddock.hi and Haddock.dyn_o in $S/dist/build/Documentation. The full /var/tmp/portage/dev-haskell/haddock-2.14.3.0.7.8.3 dir tree is attached. Interesting, you have a single .dyn_hi file there. Looks suspiciously like ghc bug Alexander pointed to. I wonder why I can't reproduce it. Let's try two tests: 1. try to reproduce with MAKEOPTS=-j1 (it will make build more deterministic) 2. try to build with HCFLAGS="-O -v" and attach build.log here. It will increase verbosity of compiler. Created attachment 392718 [details] verbose build log (In reply to Sergei Trofimovich from comment #4) Build log with HCFLAGS="-O -v" MAKEOPTS=-j1 attached. As Mark Wright says you rebuilding *ghc* with MAKEOPTS=-j1 or using USE="binary" fixes the issue, currently I'm trying to rebuild my system and understand if it will save me. I'm also experience same issue but with a different set of packages. (In reply to Alexander Vershilov from comment #6) Interesting, rebuilding ghc with MAKEOPTS=-j1 (and then rebuilding cabal and ghc-paths) did fix the problem. Unfortunately, with -j1 compiling ghc takes almost 70 minutes. *** This bug has been marked as a duplicate of bug 556504 *** |