|Version 39 (modified by jacintheford, 3 years ago) (diff)|
Getting started with the build system
There is more complete documentation for the build system later on in this guide (see Building/Using), but this page gives a quick overview of how to get around the GHC build system.
Before starting your build
You need to configure your build: which things to build, how much optimisation to use, whether to build profiling libraries, and so on. If you don't do this, then you get everything, and it will be optimised to the hilt, which means the build will take a Very Long Time. This is fine if you wanted to build GHC for installation and use, but not if you're building GHC to do some development work on it.
To configure your build, create the file mk/build.mk from the sample:
$ cp mk/build.mk.sample mk/build.mk
and then edit mk/build.mk to select the build you want. If you're not sure what you want, just uncommenting this line
#BuildFlavour = quick
Starting the build
To build the whole thing (compiler, libraries, compiler again), do this:
$ perl boot $ ./configure $ make
assuming everything goes according to plan, this should leave you with a compiler that you can run: inplace/bin/ghc-stage2.
Building after making changes
To bring the whole tree up to date after making a change, just
it might take a while, depending on what you changed. If you want to rebuild just part of the tree, for example the RTS, go into the appropriate subdirectory and say make there:
$ cd rts $ make
this should rebuild just the RTS. If you want to just build the stage 2 compiler, then
$ cd ghc $ make 2
For more, see Developing in a GHC build tree.
To clean the whole tree:
$ make clean
there's also make distclean, which will clean files that are generated by configure, and make maintainer-clean, which cleans everything that is not in the source repository.
You can clean just a part of the tree, e.g. the RTS:
$ cd rts $ make clean
The validation script does a build of GHC from scratch and runs the test suite. If it passes without any errors, then it is ok to submit the patches from your tree. You should of course either be adding onto the test suite or running manual tests to verify your changes.
The GHC build system works with make's -j flag, which spawns multiple compile processes in parallel. Even on a single processor machine it's usually worthwhile using at least make -j2, because the I/O will be overlapped with compute-intensive compilation. On a multicore machine, higher -j values will speed up the build even more.
Running GHC from the build tree
You don't need to install GHC to use it. After the build has completed, you can run GHC like this:
and to start GHCi, just add the --interactive flag. You can also see what packages have been built:
$ ./inplace/bin/ghc-pkg list
Video: Getting and Building, layout of the source tree, how to set up build.mk (23'43")