This project is mirrored from https://github.com/haskell/Cabal.
Pull mirroring updated .
- May 28, 2016
-
-
tuncer authored
-
Mikhail Glushenkov authored
bootstrap.sh bugfixes and sh compat
-
tuncer authored
1. Assign the var and export afterwards. 2. Use mktemp instead of home-grown solution. The mktemp flags we use are supported on FreeBSD, Solaris, OpenBSD, and Linux.
-
tuncer authored
The order of the 2>&1 and the redirect matters. The 2>&1 has to be last.
-
tuncer authored
Don't use variables in the printf format string. Use printf "..%s.." "$foo".
-
tuncer authored
This avoids the need to cd back.
-
tuncer authored
This catches if cd fails and also avoid the need to cd .. back.
-
tuncer authored
-
tuncer authored
Note that A && B || C is not if-then-else. C may run when A is true.
-
- May 27, 2016
-
-
Mikhail Glushenkov authored
Fix handling of default setup deps
-
Duncan Coutts authored
Covers 3 of the 4 possible cases: 1. explicit custom setup deps 2. custom setup with implicit/default deps 4. non-custom setup using the internal cabal lib version case 3 is a non-custom setup but where we're forced to use an external cabal lib version. This case is hard to test since it only happens when it's a newer (not older) Cabal lib version that the package requires, e.g. a .cabal file that specifies cabal-version: >= 2.0. Also, add a --with-ghc option to the integration test suite, which lets us more easily test with different ghc versions. Also, don't use parallel builds in any of the integration tests, as the self-exec method will not work, and some tests need to install deps for some ghc versions.
-
- May 24, 2016
-
-
Mikhail Glushenkov authored
haddock/hscolour: fix highlighted source location
-
Benno Fünfstück authored
-
Benno Fünfstück authored
When generating haddocks with the `--for-hackage` switch, the generated haddocks are placed in a different directory than the normal ones, which includes the package id instead of just the package name. When we ran hscolour, we didn't respect this, so the highlighted source would not be placed in the correct directory and thus was missing from the tarball. This patch fixes that. Fixes #3451
-
Mikhail Glushenkov authored
travis-script.sh: fix POSIX and sh compat
-
Mikhail Glushenkov authored
Updated --package-db docs
-
Andy Craze authored
There is also information on how to reset the stack. Resolve #2345
-
- May 23, 2016
-
-
tuncer authored
* remove bashisms * fix POSIX compliance * change shebang to /bin/sh
-
- May 22, 2016
-
-
Mikhail Glushenkov authored
Implement -j in travis and bootstrap script
-
Mikhail Glushenkov authored
bootstrap.sh: fix duplicate GHC_VER definition
-
- May 21, 2016
-
-
tuncer authored
-
tuncer authored
-
tuncer authored
-
Mikhail Glushenkov authored
No need to install cabal-install package
-
- May 20, 2016
-
-
Mikhail Glushenkov authored
[ci skip]
-
tuncer authored
Now that we bootstrap Cabal and cabal-install, we don't have to install the package anymore.
-
Mikhail Glushenkov authored
-
-
Mikhail Glushenkov authored
[ci skip]
-
- May 19, 2016
-
-
Mikhail Glushenkov authored
-
-
Mikhail Glushenkov authored
[ci skip]
-
GHC 7.8 renamed "Gcc Linker flags" to "C compiler link flags" and "Ld Linker flags" to "ld flags" (https://ghc.haskell.org/trac/ghc/ticket/4862 ). Without this corresponding fix in Cabal, the old-time package fails to compile on Ubuntu yakkety amd64, where the system GCC now defaults to -pie and GHC has $ ghc --info | grep -i pie ,("C compiler flags","-fno-PIE -fno-stack-protector") ,("C compiler link flags","-no-pie") ,("ld flags","-no-pie") Signed-off-by:
Anders Kaseorg <andersk@mit.edu> (cherry picked from commit c993a0c6)
-
- May 17, 2016
-
-
Mikhail Glushenkov authored
[ci skip]
-
Mikhail Glushenkov authored
-
Mikhail Glushenkov authored
-
-
Edward Z. Yang authored
Signed-off-by:
Edward Z. Yang <ezyang@cs.stanford.edu>
-
Duncan Coutts authored
Try using smaller package ids on Windows (see #3430)
-
- May 16, 2016
-
-
Duncan Coutts authored
Commit 2f976576 added default setup dep handling to the existing install command code paths, but unfortunately broke the handling for the new-build code path. It added a call to addDefaultSetupDependencies into the standardInstallPolicy. That interfered with the addDefaultSetupDependencies that ProjectPlanning was already using. So this patch splits a basicInstallPolicy out of standardInstallPolicy, where the basicInstallPolicy is all the old stuff, and the standardInstallPolicy just adds the addDefaultSetupDependencies that the install/fetch/freeze commands need. So then ProjectPlanning uses just the basicInstallPolicy. The 2f976576 commit also added a new and simpler method to determine if a package has had default setup deps added. Previously ProjectPlanning had to use a rather complex method to remember this information. So this patch removes all that and makes use of the new method. To stop this breaking in future the next patch adds integration tests to cover custom setup script handling. This fixed issue #3394.
-