No edit summary |
|||
(2 intermediate revisions by the same user not shown) | |||
Line 1: | Line 1: | ||
This guide explains in several easy steps how to get started with [[Devshell | Devshell]]. It goes through the steps of getting the source, and building a sample Haskell package in not so many not so easy steps. | |||
These directions are clearly not going to be the final workflow. If nothing else, i can't even get <code>brainfuck</code> 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. | These directions are clearly not going to be the final workflow. If nothing else, i can't even get <code>brainfuck</code> 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. | In order to begin, you will need the latest source. Currently, this is only available through git. | ||
Follow the directions in the [[Devshell development | Development Guide]] in order to get the latest upstream. | |||
A package for Fedora is currently being submitted for review. | |||
Since this is the raw upstream, there's a chance the code may not work. | Since this is the raw upstream, there's a chance the code may not work. | ||
= Launching Ports = | |||
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. | 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. | ||
Line 17: | Line 19: | ||
$ export PATH=/path/to/fedora-devshell/:$PATH | $ export PATH=/path/to/fedora-devshell/:$PATH | ||
$ ports | $ ports | ||
* PackageSource | * PackageSource | ||
* SourceBall | * SourceBall | ||
Line 59: | Line 61: | ||
* Package | * Package | ||
Traceback (most recent call last): | Traceback (most recent call last): | ||
File "/path/to/fedora-devshell/ports | File "/path/to/fedora-devshell/ports", line 36, in <module> | ||
main() | main() | ||
File "/path/to/fedora-devshell/ports | File "/path/to/fedora-devshell/ports", line 31, in main | ||
output, module, params = do_command(args) | output, module, params = do_command(args) | ||
File "/path/to/fedora-devshell/base/base.py", line 176, in do_command | File "/path/to/fedora-devshell/base/base.py", line 176, in do_command | ||
Line 71: | Line 73: | ||
This confirms that the code is running at least partially. The list of names there is a list of the available modules, albeit repeated. | 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. | 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. | ||
<pre> | <pre> | ||
$ ports | $ ports package brainfuck | ||
$ cd brainfuck/ | $ cd brainfuck/ | ||
Line 90: | Line 92: | ||
We've started a new package called brainfuck. | We've started a new package called brainfuck. | ||
<pre> | <pre> | ||
brainfuck $ ports | brainfuck $ ports haskellport add_latest brainfuck | ||
brainfuck $ ls -Al | brainfuck $ ls -Al | ||
Line 130: | Line 132: | ||
</pre> | </pre> | ||
= 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.) | 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.) | ||
Line 171: | Line 173: | ||
<pre> | <pre> | ||
brainfuck/brainfuck-0.1 $ ports | brainfuck/brainfuck-0.1 $ ports sourceball generate_patch FHS.compliance | ||
brainfuck/brainfuck-0.1 $ ls .pkg_src/ | brainfuck/brainfuck-0.1 $ ls .pkg_src/ | ||
Line 192: | Line 194: | ||
authors' reputations. | authors' reputations. | ||
brainfuck/brainfuck-0.1 $ ports | brainfuck/brainfuck-0.1 $ ports sourceball import_other_patch ../moar.FHS.patch 0 | ||
brainfuck/brainfuck-0.1 $ ls .pkg_src/ | brainfuck/brainfuck-0.1 $ ls .pkg_src/ | ||
Line 198: | Line 200: | ||
0002.moar.FHS.patch branches/ patch.log | 0002.moar.FHS.patch branches/ patch.log | ||
brainfuck/brainfuck-0.1 $ ports | brainfuck/brainfuck-0.1 $ ports sourceball clean_orig | ||
brainfuck/brainfuck-0.1 $ ports | brainfuck/brainfuck-0.1 $ ports sourceball verify_patches 'orig' | ||
[INFO] Verified clean patches: True | [INFO] Verified clean patches: True | ||
</pre> | </pre> | ||
Line 206: | Line 208: | ||
The 0 in the last command is what would ordinarily be passed to the argument <code>-p</code> to <code>patch</code>. We can also tell that the patches apply to the source code cleanly. Next we want to make a spec file. | The 0 in the last command is what would ordinarily be passed to the argument <code>-p</code> to <code>patch</code>. We can also tell that the patches apply to the source code cleanly. Next we want to make a spec file. | ||
= Creating an RPM = | |||
<pre> | <pre> | ||
brainfuck/brainfuck-0.1 $ ports | brainfuck/brainfuck-0.1 $ ports cabal . gen_spec brainfuck | ||
brainfuck/brainfuck-0.1 $ ls -A | brainfuck/brainfuck-0.1 $ ls -A | ||
Line 225: | Line 227: | ||
brainfuck $ cd .. | brainfuck $ cd .. | ||
$ ports | $ ports profile profiles/devel configure_from_system devel | ||
$ cd brainfuck/ | $ cd brainfuck/ | ||
brainfuck $ ports | brainfuck $ ports package fetch_sourceballs | ||
brainfuck $ ls -Al | brainfuck $ ls -Al | ||
Line 253: | Line 255: | ||
~/rpmbuild/SRPMS: | ~/rpmbuild/SRPMS: | ||
$ ports | $ ports build ~/rpmbuild setup_source brainfuck | ||
$ ls -l ~/rpmbuild/* | $ ls -l ~/rpmbuild/* | ||
Line 283: | Line 285: | ||
<code>mock</code> will actually run setup_sources for us later on. This just demonstrates what it does. The code is also a bit half-baked. | <code>mock</code> 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 = | |||
<pre> | <pre> | ||
$ ports | $ ports mock profiles/devel ~/rpmbuild build_rpm brainfuck | ||
[INFO] mock compiling brainfuck-0.1-1.fc10.src.rpm... please wait | [INFO] mock compiling brainfuck-0.1-1.fc10.src.rpm... please wait | ||
Line 341: | Line 343: | ||
In reality, setting up the source is not necessary. In the latest upstream, mock.build_rpm can actually set up the source for you. | In reality, setting up the source is not necessary. In the latest upstream, mock.build_rpm can actually set up the source for you. | ||
= Conclusion = | |||
Clearly <code>mock</code> 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 <code>mock</code>. Fedora Devshell will automatically build the source RPM, and send the appropriate command to <code>mock</code> to use a premade profile. There will be more you can do with this here. | Clearly <code>mock</code> 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 <code>mock</code>. Fedora Devshell will automatically build the source RPM, and send the appropriate command to <code>mock</code> to use a premade profile. There will be more you can do with this here. |
Latest revision as of 06:10, 30 April 2009
This guide explains in several easy steps how to get started with Devshell. It goes through the steps of getting the source, and building a sample Haskell package in not so many not so easy steps.
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.
Follow the directions in the Development Guide in order to get the latest upstream.
A package for Fedora is currently being submitted for review.
Since this is the raw upstream, there's a chance the code may not work.
Launching Ports
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 * 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", line 36, in <module> main() File "/path/to/fedora-devshell/ports", 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 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 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 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 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 sourceball clean_orig brainfuck/brainfuck-0.1 $ ports sourceball verify_patches 'orig' [INFO] Verified clean patches: True
The 0 in the last command is what would ordinarily be passed to the argument -p
to 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 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. haskellports
uses 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 mock
brainfuck $ cd .. $ ports profile profiles/devel configure_from_system devel $ cd brainfuck/ brainfuck $ ports 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 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 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.
Conclusion
Clearly 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.