These directions are clearly not going to be the final workflow. If nothing else, i can't even get
brainfuck to compile fully. But it demonstrates fundamentally how the workflow will follow. Most of the steps in this process are pretty generic to alot of kinds of packages. Implementing code to handle the bits for other packages is as simple as coding the steps into Python.
Getting the Source
In order to begin, you will need the latest source. Currently, this is only available through git. Open a handy terminal window and follow along.
$ git clone git://git.fedorapeople.org/home/fedora/ynemoy/public_git/fedora-devshell.git
Since this is the raw upstream, there's a chance the code may not work.
Since there are no built in tools yet for building a package, the only way to interact with the code is via the source directory itself. Therefore it's necessary to modify the command line path to make this work.
$ export PATH=/path/to/fedora-devshell/:$PATH $ ports.py * PackageSource * SourceBall * Audit * Directory * Profile * Directory * BuildSystem * Darcs * Hackage * HaskellPort * Port * SourceBall * Darcs * RevisionControl * Build * Directory * Package * Profile * BuildSystem * Cabal * SourceBall * Directory * Repo * Directory * PackageSource * Bugs * Mail * Source * Build * Mock * Package * Profile * Port * PackageSource * RevisionControl * Fetcher * Fetcher * Hackage * Directory * Package Traceback (most recent call last): File "/path/to/fedora-devshell/ports.py", line 36, in <module> main() File "/path/to/fedora-devshell/ports.py", line 31, in main output, module, params = do_command(args) File "/path/to/fedora-devshell/base/base.py", line 176, in do_command output = top(*params) TypeError: 'NoneType' object is not callable
This confirms that the code is running at least partially. The list of names there is a list of the available modules, albeit repeated.
Setting up a project
Currently, Fedora Devshell can only work with Haskell programs, especially ones hosted on Hackage, the Haskell package repository. It also only works with packages managed by Darcs, the Haskell distributed version control system. Let's get a package and start playing with it.
$ ports.py package brainfuck $ cd brainfuck/ brainfuck $ ls -Al totaal 4 -rw-r--r-- 1 yankee yankee 44 jan 20 22:19 .devshell brainfuck $ cat .devshell type = package name = brainfuck sources = ,
We've started a new package called brainfuck.
brainfuck $ ports.py haskellport add_latest brainfuck brainfuck $ ls -Al totaal 8 drwxr-sr-x 4 yankee yankee 4096 jan 20 22:24 brainfuck-0.1/ -rw-r--r-- 1 yankee yankee 57 jan 20 22:24 .devshell brainfuck $ cd brainfuck-0.1/ brainfuck/brainfuck-0.1 $ ls -Al totaal 44 -rw------- 1 yankee yankee 852 jun 8 2008 brainfuck.cabal -rw-r--r-- 1 yankee yankee 200 jan 20 22:24 .devshell drwx------ 3 yankee yankee 4096 jun 8 2008 Language/ -rw------- 1 yankee yankee 17992 jun 8 2008 LICENSE -rw------- 1 yankee yankee 439 jun 8 2008 Main.hs drwxr-sr-x 3 yankee yankee 4096 jan 20 22:24 .pkg_src/ -rw------- 1 yankee yankee 95 jun 8 2008 Setup.hs brainfuck/brainfuck-0.1 $ cat .devshell type = sourceball name = brainfuck-0.1 tarball_source = http://hackage.haskell.org/packages/archive/brainfuck/0.1/brainfuck-0.1.tar.gz sourceball = brainfuck-0.1.tar.gz source = . buildsystem = cabal brainfuck/brainfuck-0.1 $ ls .pkg_src/ brainfuck-0.1.tar.gz branches/ yankee@koan ~/haskell/brainfuck/brainfuck-0.1 $ ls .pkg_src/bra brainfuck-0.1.tar.gz branches/ yankee@koan ~/haskell/brainfuck/brainfuck-0.1 $ ls .pkg_src/branches/ brainfuck-0.1_orig/ yankee@koan ~/haskell/brainfuck/brainfuck-0.1 $ ls .pkg_src/branches/brainfuck-0.1_orig/ brainfuck.cabal Language/ LICENSE Main.hs Setup.hs
Editing the package
This gets the latest upstream tarball from hackage. Let's do a bit of hacking. Let's assume that brainfuck's not FHS compliant, and we need to put a quick patch in to get it into Fedora. Eventually we will send that patch upstream, but the tools for that are still in the works. For now, i'm going to assume that adding lines to the LICENSE file will make it FHS compliant. (In my imaginary world, we also have cookies.)
brainfuck/brainfuck-0.1 $ vi LICENSE brainfuck/brainfuck-0.1 $ diff -ru -X /path/to/fedora-devshell/diff.excludes .pkg_src/branches/brainfuck-0.1_orig/ . diff -ru -X /path/to/fedora-devshell/diff.excludes .pkg_src/branches/brainfuck-0.1_orig/LICENSE ./LICENSE --- .pkg_src/branches/brainfuck-0.1_orig/LICENSE 2008-06-08 18:37:11.000000000 -0400 +++ ./LICENSE 2009-01-20 22:29:40.911210433 -0500 @@ -26,6 +26,8 @@ in new free programs; and that you know you can do these things. To protect your rights, we need to make restrictions that forbid + + anyone to deny you these rights or to ask you to surrender the rights. These restrictions translate to certain responsibilities for you if you distribute copies of the software, or if you modify it. @@ -33,6 +35,8 @@ For example, if you distribute copies of such a program, whether gratis or for a fee, you must give the recipients all the rights that you have. You must make sure that they, too, receive or can get the + + source code. And you must show them these terms so they know their rights. @@ -43,6 +47,7 @@ Also, for each author's protection and ours, we want to make certain that everyone understands that there is no warranty for this free software. If the software is modified by someone else and passed on, we +RMS has a big beard. want its recipients to know that what they have is not the original, so that any problems introduced by others will not reflect on the original authors' reputations.
As you can see, keeping the original code tree around is convenient, because we can run diff's easily.
brainfuck/brainfuck-0.1 $ ports.py sourceball generate_patch FHS.compliance brainfuck/brainfuck-0.1 $ ls .pkg_src/ 0001.FHS.compliance.patch brainfuck-0.1.tar.gz branches/ diff.log patch.log
This creates our first patch. It's ordered for reasons that will be obvious shortly. Later in the workflow, someone hands you a second patch, which you want to import into the package. Apparently you couldn't make the package FHS compliant enough.
brainfuck/brainfuck-0.1 $ cat ../moar.FHS.patch diff -ru -X /path/to/fedora-devshell/diff.excludes .pkg_src/branches/brainfuck-0.1_orig/LICENSE ./LICENSE --- .pkg_src/branches/brainfuck-0.1_orig/LICENSE 2009-01-20 22:37:21.186153827 -0500 +++ ./LICENSE 2009-01-20 22:36:21.964170098 -0500 @@ -48,6 +48,7 @@ that everyone understands that there is no warranty for this free software. If the software is modified by someone else and passed on, we RMS has a big beard. +It's probably going gray. want its recipients to know that what they have is not the original, so that any problems introduced by others will not reflect on the original authors' reputations. brainfuck/brainfuck-0.1 $ ports.py sourceball import_other_patch ../moar.FHS.patch 0 brainfuck/brainfuck-0.1 $ ls .pkg_src/ 0001.FHS.compliance.patch brainfuck-0.1.tar.gz diff.log 0002.moar.FHS.patch branches/ patch.log brainfuck/brainfuck-0.1 $ ports.py sourceball clean_orig brainfuck/brainfuck-0.1 $ ports.py sourceball verify_patches 'orig' [INFO] Verified clean patches: True
The 0 in the last command is what would ordinarily be passed to the argument
patch. We can also tell that the patches apply to the source code cleanly. Next we want to make a spec file.
Creating an RPM
brainfuck/brainfuck-0.1 $ ports.py cabal . gen_spec brainfuck brainfuck/brainfuck-0.1 $ ls -A brainfuck.cabal .devshell Language/ Main.hs Setup.hs brainfuck.spec .devshell~ LICENSE .pkg_src/ brainfuck/brainfuck-0.1 $ cd .. brainfuck $ ls -A brainfuck-0.1/ brainfuck.spec .devshell moar.FHS.patch
I've skipped the steps about integrating patches into the spec file, and updating it with all the vital bits.
cabal2spec which means you have to fill in all the little bits later. Rather, we're going to jump right away to compiling RPMs. Currently, Haskell packages can only be built for rawhide, so we're going to use the 'devel' branch.
profile will copy over mock configuration files, so we can tinker with them later. This segment is only half baked, so theres's not much you can do with profile except copy bits and use it with
brainfuck $ cd .. $ ports.py profile profiles/devel configure_from_system devel $ cd brainfuck/ brainfuck $ ports.py package fetch_sourceballs brainfuck $ ls -Al totaal 20 drwxr-sr-x 4 yankee yankee 4096 jan 20 23:07 brainfuck-0.1/ lrwxrwxrwx 1 yankee yankee 74 jan 20 23:17 brainfuck-0.1.tar.gz -> ~/haskell/brainfuck/brainfuck-0.1/.pkg_src/brainfuck-0.1.tar.gz -rw-r--r-- 1 yankee yankee 2456 jan 20 23:11 brainfuck.spec -rw-r--r-- 1 yankee yankee 115 jan 20 23:17 .devshell -rw-r--r-- 1 yankee yankee 652 jan 20 22:37 moar.FHS.patch $ ls ~/rpmbuild/* ~/rpmbuild/BUILD: ~/rpmbuild/BUILDROOT: ~/rpmbuild/RPMS: i386/ noarch/ x86_64/ ~/rpmbuild/SOURCES: ~/rpmbuild/SPECS: ~/rpmbuild/SRPMS: $ ports.py build ~/rpmbuild setup_source brainfuck $ ls -l ~/rpmbuild/* ~/rpmbuild/BUILD: totaal 0 ~/rpmbuild/BUILDROOT: totaal 0 ~/rpmbuild/RPMS: totaal 12 drwxr-sr-x 2 yankee yankee 4096 dec 16 00:22 i386/ drwxr-sr-x 2 yankee yankee 4096 dec 16 00:22 noarch/ drwxr-sr-x 2 yankee yankee 4096 jan 7 22:50 x86_64/ ~/rpmbuild/SOURCES: totaal 0 lrwxrwxrwx 1 yankee yankee 51 jan 20 23:19 brainfuck-0.1.tar.gz -> ~/haskell/brainfuck/brainfuck-0.1.tar.gz lrwxrwxrwx 1 yankee yankee 45 jan 20 23:19 moar.FHS.patch -> ~/haskell/brainfuck/moar.FHS.patch ~/rpmbuild/SPECS: totaal 0 lrwxrwxrwx 1 yankee yankee 45 jan 20 23:19 brainfuck.spec -> ~/haskell/brainfuck/brainfuck.spec ~/rpmbuild/SRPMS: totaal 0
mock will actually run setup_sources for us later on. This just demonstrates what it does. The code is also a bit half-baked.
Building the RPM in Mock
$ ports.py mock profiles/devel ~/rpmbuild build_rpm brainfuck [INFO] mock compiling brainfuck-0.1-1.fc10.src.rpm... please wait $ cd profiles/devel/ profiles/devel $ ls -Al totaal 92 -rw-r--r-- 1 yankee yankee 15078 jan 20 23:22 brainfuck-0.1-1.fc10.src.rpm -rw-rw-r-- 1 yankee yankee 24873 jan 20 23:35 build.log -rw-r--r-- 1 yankee yankee 117 jan 20 23:35 .devshell drwxr-xr-x 2 yankee yankee 4096 jan 5 15:12 mock/ -rw-r--r-- 1 yankee yankee 1013 jan 20 23:35 mock.log -rw-rw-r-- 1 yankee yankee 24794 jan 20 23:35 root.log -rw-rw-r-- 1 yankee yankee 375 jan 20 23:35 state.log profiles/devel $ cat .devshell type = profile name = devel branch = devel koji_target = dist-devel distval = rawhide distvar = fedora dist = .devel profiles/devel $ cat mock.log -- Beginning log of mock.log at 2009-01-20 23:22:12.783201 -- INFO: mock.py version 0.9.13 starting... State Changed: init plugins State Changed: start INFO: Start(brainfuck-0.1-1.fc10.src.rpm) Config(fedora-rawhide-i386) State Changed: lock buildroot State Changed: clean State Changed: init State Changed: lock buildroot INFO: enabled root cache State Changed: unpacking root cache INFO: enabled yum cache State Changed: cleaning yum metadata INFO: enabled ccache State Changed: running yum State Changed: setup State Changed: build ERROR: Exception(brainfuck-0.1-1.fc10.src.rpm) Config(fedora-rawhide-i386) 12 minutes 58 seconds INFO: Results and/or logs in: ~/haskell/profiles/devel INFO: Cleaning up build root ('clean_on_failure=True') State Changed: lock buildroot State Changed: clean ERROR: Command failed. See logs for output. # ['bash', '--login', '-c', 'rpmbuild -bb --target i386 --nodeps builddir/build/SPECS/brainfuck.spec'] -- Ending log of mock.log at 2009-01-20 23:35:17.194359 --
In reality, setting up the source is not necessary. In the latest upstream, mock.build_rpm can actually set up the source for you.
mock failed. I guess that spec file wasn't perfect. Because of the use of symlinks, however, i can simply edit the spec file, make the necessary changes, and rerun the command to initialize
mock. Fedora Devshell will automatically build the source RPM, and send the appropriate command to
mock to use a premade profile. There will be more you can do with this here.
This concludes the tutorial section. For more information about what you can do with Fedora Devshell, have a look at the Developers Guide.