March 2011 - cabal-devel

[Hackage] #667: Improve tools to manage ad-hoc package releases
by Hackage 05 Jan '12

05 Jan '12

#667: Improve tools to manage ad-hoc package releases ------------------------------+--------------------------------------------- Reporter: duncan | Owner: Type: enhancement | Status: new Priority: normal | Milestone: Component: miscellaneous | Version: Severity: normal | Keywords: Difficulty: project(> week) | Ghcversion: Platform: | ------------------------------+--------------------------------------------- Hackage is suited for major software releases by package maintainers. It is not suited to distributing betas, daily snapshots or random forks, or patches of existing releases. For example a user may have some patch or bugfix for a package that they need. It takes time for maintainers to review and integrate these. Sometimes maintainers have little time, or they disagree with a patch approach or quality. It is right that maintainers have control over what is released on hackage. However there is a need for decentralised, ad-hoc and short term releases. We do not want to encourage the pollution of the package namespace with lots of short-term forks. One solution is to let users post package forks on their own webspace. We would want to extend cabal so that people can run: {{{ cabal install http://code.haskell.org/~user/foo/foo-fork-to-fix- blah-1.0.1.tar.gz }}} This makes it very decentralised. We may want to add some feature to hackage package pages to link to third-party forks. In addition to hosting and installing single packages it would be useful to be able to publish package collections. Of course this is exactly what a hackage archive is. We should provide easy tools to make and manage hackage archives that can be hosted on dumb http file servers. Going one step further, the hackage index format should be extended to make it more RESTful by including links to tarballs rather than just assuming that tarballs are found relative to the index file using some fixed layout. This would allow people to publish indexes that span multiple servers. This gets close to the concept of a package "search path" promoted by the searchpath project (http://searchpath.org/) Additonally, an index should allow us to point to source control repos rather than just to release tarballs. This kind of index would be useful for nighly builds. -- Ticket URL: <http://hackage.haskell.org/trac/hackage/ticket/667> Hackage <http://haskell.org/cabal/> Hackage: Cabal and related projects

1 2

[Hackage] #724: Add getDataDirectory function to Paths_pkgname
by Hackage 05 Jan '12

05 Jan '12

#724: Add getDataDirectory function to Paths_pkgname ----------------------------+----------------------------------------------- Reporter: tommd | Owner: Type: enhancement | Status: new Priority: normal | Milestone: Component: Cabal library | Version: 1.6.0.3 Severity: normal | Keywords: Difficulty: unknown | Ghcversion: Platform: | ----------------------------+----------------------------------------------- Some libraries and programs have a large number of files. Instead of having explicit knowledge of each file (and instead of using a mutation on the result of getDataFilename) there should be a "getDataDirectory" or a "getDataDirectoryContents": getDataDirectoryContents = getDataDirectory >>= getDirectoryContents -- Ticket URL: <http://hackage.haskell.org/trac/hackage/ticket/724> Hackage <http://haskell.org/cabal/> Hackage: Cabal and related projects

1 4

[Hackage] #739: cabal install regression with GHC HEAD - ghc-paths
by Hackage 05 Jan '12

05 Jan '12

#739: cabal install regression with GHC HEAD - ghc-paths ----------------------------+----------------------------------------------- Reporter: AntoineLatter | Owner: Type: defect | Status: new Priority: normal | Milestone: Component: Cabal library | Version: HEAD Severity: normal | Keywords: Difficulty: unknown | Ghcversion: Platform: | ----------------------------+----------------------------------------------- with the Cabal library installed with GHC 6.13.20100911 I can no longer install the package ghc-paths-0.1.0.6 In 'ghci Setup.hs' I get the error: Setup.hs:41:25: Ambiguous occurrence `None' It could refer to either `Distribution.Simple.Setup.None', imported from Distribution.Simple.Setup at Setup.hs:2:1-32 or `Pkg.None', imported from Distribution.Simple.PackageIndex at Setup.hs:7:1-46 It looks like the 'None' constructor didn't exist in Distribution.Simple.Setup in Cabal 1.8.x. While we can't guarantee that an added constructor doesn't conflict with data declarations used in every Setup.hs, it seems in poor form to export the same constructor in two types in the same library. -- Ticket URL: <http://hackage.haskell.org/trac/hackage/ticket/739> Hackage <http://haskell.org/cabal/> Hackage: Cabal and related projects

1 4

[Hackage] #671: better error message for broken build tools
by Hackage 05 Jan '12

05 Jan '12

#671: better error message for broken build tools ------------------------------+--------------------------------------------- Reporter: duncan | Owner: Type: defect | Status: new Priority: normal | Milestone: Component: Cabal library | Version: 1.8.0.4 Severity: normal | Keywords: Difficulty: easy (<4 hours) | Ghcversion: Platform: | ------------------------------+--------------------------------------------- Consider: {{{ Configuring haskell-src-exts-1.9.0... setup: happy version >=1.17 is required but the version of /usr/local/bin/happy could not be determined. }}} The actual problem is: {{{ /usr/local/bin/happy: error while loading shared libraries: libgmp.so.3: cannot open shared object file: No such file or directory }}} We could produce a better error message in this situation. If we get a non-0 exit code when we check the `--version` then we can probably safely report that the program is probably broken. Of course if we get some result text back and we cannot extract and parse the version number then that's different. -- Ticket URL: <http://hackage.haskell.org/trac/hackage/ticket/671> Hackage <http://haskell.org/cabal/> Hackage: Cabal and related projects

1 2

[Hackage] #663: Setup.hs doesn't look into ~/.cabal/bin for build tools.
by Hackage 05 Jan '12

05 Jan '12

#663: Setup.hs doesn't look into ~/.cabal/bin for build tools. ----------------------------------+----------------------------------------- Reporter: Axel Simon | Owner: Type: defect | Status: new Priority: normal | Milestone: Component: Cabal library | Version: 1.6.0.3 Severity: normal | Keywords: Difficulty: very easy (<1 hour) | Ghcversion: Platform: | ----------------------------------+----------------------------------------- Cabal currently only checks the current PATH for buildtools. If the user installs and new pre-processor with the default --user installation, these tools wind up in ~/.cabal/bin. However, cabal doesn't find these unless ~/.cabal/bin has been added to the PATH. This is irritating since it is then not possible to make a library depend on a cabal package that contains it's build tools: Even if cabal-install installs the build tools first, building the library still fails since the tools cannot be found. The solution is very simple: Add ~/.cabal/bin to the search path when looking for build tools! -- Ticket URL: <http://hackage.haskell.org/trac/hackage/ticket/663> Hackage <http://haskell.org/cabal/> Hackage: Cabal and related projects

1 4

[Hackage] #701: Help packages to avoid linker symbol clashes
by Hackage 05 Jan '12

05 Jan '12

#701: Help packages to avoid linker symbol clashes ----------------------------+----------------------------------------------- Reporter: duncan | Owner: Type: enhancement | Status: new Priority: normal | Milestone: Component: Cabal library | Version: 1.6.0.3 Severity: normal | Keywords: Difficulty: unknown | Ghcversion: Platform: | ----------------------------+----------------------------------------------- Currently it is all too easy to end up with silently clashing linker symbols. The prime example is C code linked into a Haskell package. Multiple versions of the same package will define the same linker symbols. The system linker is happy to just pick the first occurrence of a symbol (though the ghci linker is stricter). This can cause havoc if the behaviour of the C code changes from one version of the package to another (see for example the problem with mmap and darcs, ticket #700). There are a number of things we could do to improve the situation. * We could make it easier to make versioned C symbols in the C code used in a Haskell package by providing cpp macros. * We could explicitly check for unversioned C symbols in Haskell libs * We could try linking Haskell libs in such a way that the C symbols are not visible and thus cannot clash, but it would also mean they are not exported for other libs to use. -- Ticket URL: <http://hackage.haskell.org/trac/hackage/ticket/701> Hackage <http://haskell.org/cabal/> Hackage: Cabal and related projects

1 2

[Hackage] #803: cabal test after cabal configure --enable-tests without cabal build creates obscure output
by Hackage 05 Jan '12

05 Jan '12

#803: cabal test after cabal configure --enable-tests without cabal build creates obscure output ----------------------------+----------------------------------------------- Reporter: guest | Owner: Type: defect | Status: new Priority: normal | Milestone: Component: Cabal library | Version: 1.8.0.6 Severity: normal | Keywords: Difficulty: unknown | Ghcversion: Platform: | ----------------------------+----------------------------------------------- It wasn't obvious to me that cabal build would be required before cabal test. Can cabal test automatically run cabal build? Otherwise a message should be shown telling the user to run cabal build. Here is an example output {{{ >>> Test suite runtests: RUNNING... >>> Test suite runtests: FAIL >>> Test suite logged to: dist/test/persistent-0.4.0.1-runtests.log >>> Test suite runtests: RUNNING... >>> Test suite runtests: FAIL >>> Test suite logged to: dist/test/persistent-0.4.0.1-runtests.log Test suite runtests: FAIL Test suite logged to: dist/test/persistent-0.4.0.1-runtests.log 0 of 1 test suites (0 of 1 test cases) passed. }}} The log is exactly the same as the output -- Ticket URL: <http://hackage.haskell.org/trac/hackage/ticket/803> Hackage <http://haskell.org/cabal/> Hackage: Cabal and related projects

1 4

[Hackage] #814: Avoid orphan flexible instances in Cabal-TestSuite documentation
by Hackage 05 Jan '12

05 Jan '12

#814: Avoid orphan flexible instances in Cabal-TestSuite documentation ----------------------------------+----------------------------------------- Reporter: guest | Owner: Type: enhancement | Status: new Priority: normal | Milestone: Component: User guide | Version: 1.10.1.0 Severity: normal | Keywords: Difficulty: very easy (<1 hour) | Ghcversion: Platform: | ----------------------------------+----------------------------------------- The example code shown in http://www.haskell.org/cabal/release/cabal-1.10.1.0/doc/users-guide/#test- suites needs FlexibleInstances, and the instance is orphan, too. Although this code is not intended to be used in most packages, it still looks like those instances are necessary or at least good style. I suggest to replace the generic pair by a custom data type like {{{ data TestCase = TestCase String Bool instance TestOptions TestCase where ... }}} in the documentation. Since Cabal is a widely used library, it has automatically an educational aspect. -- Ticket URL: <http://hackage.haskell.org/trac/hackage/ticket/814> Hackage <http://haskell.org/cabal/> Hackage: Cabal and related projects

1 3

[Hackage] #766: hackage build logs gone missing?
by Hackage 05 Jan '12

05 Jan '12

#766: hackage build logs gone missing? --------------------------------+------------------------------------------- Reporter: claus | Owner: Type: defect | Status: new Priority: normal | Milestone: Component: hackageDB website | Version: 1.8.0.6 Severity: normal | Keywords: Difficulty: unknown | Ghcversion: Platform: | --------------------------------+------------------------------------------- It seems that build logs are no longer linked from package pages (used to be linked from the `built on` entries, iirc), such as: http://hackage.haskell.org/package/haddock Also, the `latest/` directory has `doc/`, but not `logs/`, while the archived versioned directory has both: http://hackage.haskell.org/packages/archive/haddock/latest/ http://hackage.haskell.org/packages/archive/haddock/2.8.1/ And, since I've more than once confused myself with that directory structure (`package` singular, but `packages/archive` plural? why the different prefix, where a version number postfix should be enough? why both `http://hackage.haskell.org/package/haddock-2.8.0` and `http://hackage.haskell.org/packages/archive/haddock/2.8.0/`?), could that be simplified (or, at least, a link be provided from the `package`-entry to the shadow-hierarchy `packages/archive`)? -- Ticket URL: <http://hackage.haskell.org/trac/hackage/ticket/766> Hackage <http://haskell.org/cabal/> Hackage: Cabal and related projects

1 4

[Hackage] #791: Large libraries not installable on OS X
by Hackage 05 Jan '12

05 Jan '12

#791: Large libraries not installable on OS X -----------------------------+---------------------------------------------- Reporter: batterseapower | Owner: Type: defect | Status: new Priority: high | Milestone: Component: Cabal library | Version: 1.8.0.6 Severity: normal | Keywords: Difficulty: unknown | Ghcversion: Platform: Mac OS | -----------------------------+---------------------------------------------- For example: {{{ cabal unpack highlighting-kate-0.2.8.1 cd highlighting-kate-0.2.8.1 ghc --make Setup.hs ./Setup configure ./Setup build }}} This fails: {{{ ld: scattered reloc r_address too large for inferred architecture i386 }}} The underlying reason is that Cabal is trying to create an object file for use with GHCi, but that is not possible (see http://hackage.haskell.org/trac/ghc/ticket/3260) Cabal's GHC.hs should not use combineObjectFiles on with GHC >= 7 and OS X. -- Ticket URL: <http://hackage.haskell.org/trac/hackage/ticket/791> Hackage <http://haskell.org/cabal/> Hackage: Cabal and related projects

4 15