When we release versions of Privoxy, our work leaves our cozy secret lab and has to work in the cold RealWorld[tm]. Once it is released, there is no way to call it back, so it is very important that great care is taken to ensure that everything runs fine, and not to introduce problems in the very last minute.
So when releasing a new version, please adhere exactly to the procedure outlined in this chapter.
The following programs are required to follow this process: ncftpput (ncftp), scp, ssh (ssh), gmake (GNU's version of make), autoconf, cvs.
First you need to determine which version number the release will have. Privoxy version numbers consist of three numbers, separated by dots, like in X.Y.Z (e.g. 3.0.0), where:
X, the version major, is rarely ever changed. It is increased by one if turning a development branch into stable substantially changes the functionality, user interface or configuration syntax. Majors 1 and 2 were Junkbuster, and 3 will be the first stable Privoxy release.
Y, the version minor, represents the branch within the major version. At any point in time, there are two branches being maintained: The stable branch, with an even minor, say, 2N, in which no functionality is being added and only bug-fixes are made, and 2N+1, the development branch, in which the further development of Privoxy takes place. This enables us to turn the code upside down and inside out, while at the same time providing and maintaining a stable version. The minor is reset to zero (and one) when the major is incremented. When a development branch has matured to the point where it can be turned into stable, the old stable branch 2N is given up (i.e. no longer maintained), the former development branch 2N+1 becomes the new stable branch 2N+2, and a new development branch 2N+3 is opened.
Z, the point or sub version, represents a release of the software within a branch. It is therefore incremented immediately after each software release. The point version is reset to zero when the minor changes.
Stable branches work a little differently, since there should be little to no development happening in such branches. Remember, only bugfixes, which presumably should have had some testing before being committed. Stable branches will then have their version reported as 0.0.0, during that period between releases when changes are being added. This is to denote that this code is not for release. Then as the release nears, the version is bumped according: e.g. 3.0.1 -> 0.0.0 -> 3.0.2.
In summary, the main Git trunk is the development branch where new features are being worked on for the next stable series. This should almost always be where the most activity takes place. There is always at least one stable branch from the trunk, e.g now it is 3.0, which is only used to release stable versions. Once the initial *.0 release of the stable branch has been done, then as a rule, only bugfixes that have had prior testing should be committed to the stable branch. Once there are enough bugfixes to justify a new release, the version of this branch is again incremented Example: 3.0.0 -> 3.0.1 -> 3.0.2, etc are all stable releases from within the stable branch. 3.1.x is currently the main trunk, and where work on 3.2.x is taking place. If any questions, please post to the devel list before committing to a stable branch!
Developers should remember too that if they commit a bugfix to the stable branch, this will more than likely require a separate submission to the main trunk, since these are separate development trees within Git. If you are working on both, then this would require at least two separate check outs (i.e main trunk, and the stable release branch, which is v_3_0_branch at the moment).
The following must be done by one of the developers prior to each new release.
Make sure that everybody who has worked on the code in the last couple of days has had a chance to yell "no!" in case they have pending changes/fixes in their pipelines. Announce the freeze so that nobody will interfere with last minute changes.
Update the code status (CODE_STATUS="xxx") in configure.in to one of "alpha", "beta" or "stable".
Rebuild configure and GNUMakefile to make sure the updated values are being used.
$ autoheader && autoconf # rebuild configure $ ./configure # rebuild GNUmakefile |
make dok-release to update the sgml documentation source files.
If action file processing has changed and is not backward-compatible, make sure the "for-privoxy-version=x.y.z" minimum version number in default.action.master has been updated:
{{settings}} ############################################################################# #MASTER# COMMENT: The minimum Privoxy version: for-privoxy-version=3.0.11 |
Create the change log:
$ git tag # to see the tags $ git log [last release tag]..HEAD > /tmp/log # get the commit log since the last release $ utils/makeChangeLog /tmp/log > /tmp/change.log # reformat the commit log |
Edit /tmp/change.log to remove trivial changes and group the changes under general headings like:
- Bug fixes: - Action file improvements: - Filter file improvements: - General improvements: - Documentation improvements: - Build system improvements: - Code cleanups: - Privoxy-Log-Parser: - Privoxy-Regression-Test: |
Add the contents of /tmp/change.log to the start of ChangeLog and re-create doc/source/changelog.sgml:
$ utils/changelog2doc.pl /tmp/change.log >| doc/source/changelog.sgml |
All developers should look at the ChangeLog and make sure noteworthy changes are referenced.
All documentation should be rebuilt:
$ make dok $ make dok-tidy $ make man $ make config-file |
Commit all files that were changed in the above steps!
The User Manual is also used for context sensitive help for the CGI editor. This is version sensitive, so that the user will get appropriate help for his/her release. So with each release a fresh version should be uploaded to the webserver (this is in addition to the main User Manual link from the main page since we need to keep manuals for various versions available). The CGI pages will link to something like http://privoxy.org/$(VERSION)/user-manual/. This will need to be updated for each new release. There is no Makefile target for this at this time!!! It needs to be done manually.
Tag all files in Git with the version number with "cvs tag v_X_Y_Z". Don't use vX_Y_Z, ver_X_Y_Z, v_X.Y.Z (won't work) etc.
On the webserver, copy the user manual to a new top-level directory called X.Y.Z. This ensures that help links from the CGI pages, which have the version as a prefix, will go into the right version of the manual. If this is a development branch release, also symlink X.Y.(Z-1) to X.Y.Z and X.Y.(Z+1) to . (i.e. dot).
Now the individual packages can be built and released. Note that for GPL reasons the first package to be released is always the source tarball.
For all types of packages, including the source tarball, you must make sure that you build from clean sources by exporting the right version from Git into an empty directory (just press return when asked for a password):
mkdir dist # delete or choose different name if it already exists cd dist cvs -d:pserver:anonymous@ijbswa.cvs.sourceforge.net:/cvsroot/ijbswa login cvs -z3 -d:pserver:anonymous@ijbswa.cvs.sourceforge.net:/cvsroot/ijbswa export -r v_X_Y_Z current |
Do NOT change a single bit, including, but not limited to version information after export from Git. This is to make sure that all release packages, and with them, all future bug reports, are based on exactly the same code.
Warning |
Every significant release of Privoxy has included at least one package that either had incorrect versions of files, missing files, or incidental leftovers from a previous build process that gave unknown numbers of users headaches to try to figure out what was wrong. PLEASE, make sure you are using pristene sources, and are following the prescribed process! |
Please find additional instructions for the source tarball and the individual platform dependent binary packages below. And details on the Sourceforge release process below that.
Please keep these general guidelines in mind when putting together your package. These apply to all platforms!
Privoxy requires write access to: all *.action files, all logfiles, and the trust file. You will need to determine the best way to do this for your platform.
Please include up to date documentation. At a bare minimum:
LICENSE (top-level directory) |
README (top-level directory) |
AUTHORS (top-level directory) |
man page (top-level directory, Unix-like platforms only) |
The User Manual (doc/webserver/user-manual/) |
FAQ (doc/webserver/faq/) |
Also suggested: Developer Manual (doc/webserver/developer-manual) and ChangeLog (top-level directory). FAQ and the manuals are HTML docs. There are also text versions in doc/text/ which could conceivably also be included.
The documentation has been designed such that the manuals are linked to each other from parallel directories, and should be packaged that way. privoxy-index.html can also be included and can serve as a focal point for docs and other links of interest (and possibly renamed to index.html). This should be one level up from the manuals. There is a link also on this page to an HTMLized version of the man page. To avoid 404 for this, it is in Git as doc/webserver/man-page/privoxy-man-page.html, and should be included along with the manuals. There is also a css stylesheets that can be included for better presentation: p_doc.css. This should be in the same directory with privoxy-index.html, (i.e. one level up from the manual directories).
user.action and user.filter are designed for local preferences. Make sure these do not get overwritten! config should not be overwritten either. This has especially important configuration data in it. trust should be left in tact as well.
Other configuration files (default.action and default.filter) should be installed as the new defaults, but all previously installed configuration files should be preserved as backups. This is just good manners :-) These files are likely to change between releases and contain important new features and bug fixes.
Please check platform specific notes in this doc, if you haven't done "Privoxy" packaging before for other platform specific issues. Conversely, please add any notes that you know are important for your platform (or contact one of the doc maintainers to do this if you can't).
Packagers should do a "clean" install of their package after building it. So any previous installs should be removed first to ensure the integrity of the newly built package. Then run the package for a while to make sure there are no obvious problems, before uploading.
First, make sure that you have freshly exported the right version into an empty directory. (See "Building and releasing packages" above). Then run:
cd current autoheader && autoconf && ./configure |
Then do:
make tarball-dist |
To upload the package to Sourceforge, simply issue
make tarball-upload |
Go to the displayed URL and release the file publicly on Sourceforge. For the change log field, use the relevant section of the ChangeLog file.
In following text, replace dist with either "rh" for Red Hat or "suse" for SuSE.
First, make sure that you have freshly exported the right version into an empty directory. (See "Building and releasing packages" above).
As the only exception to not changing anything after export from Git, now examine the file privoxy-dist.spec and make sure that the version information and the RPM release number are correct. The RPM release numbers for each version start at one. Hence it must be reset to one if this is the first RPM for dist which is built from version X.Y.Z. Check the file list if unsure. Else, it must be set to the highest already available RPM release number for that version plus one.
Then run:
cd current autoheader && autoconf && ./configure |
Then do
make dist-dist |
To upload the package to Sourceforge, simply issue
make dist-upload rpm_packagerev |
where rpm_packagerev is the RPM release number as determined above. Go to the displayed URL and release the file publicly on Sourceforge. Use the release notes and change log from the source tarball package.
First, make sure that you have freshly exported the right version into an empty directory. (See "Building and releasing packages" above). Then get the OS/2 Setup module:
cvs -z3 -d:pserver:anonymous@ijbswa.cvs.sourceforge.net:/cvsroot/ijbswa co os2setup |
You will need a mix of development tools. The main compilation takes place with IBM Visual Age C++. Some ancillary work takes place with GNU tools, available from various sources like hobbes.nmsu.edu. Specificially, you will need autoheader, autoconf and sh tools. The packaging takes place with WarpIN, available from various sources, including its home page: xworkplace.
Change directory to the os2setup directory. Edit the os2build.cmd file to set the final executable filename. For example,
installExeName='privoxyos2_setup_X.Y.Z.exe' |
Next, edit the IJB.wis file so the release number matches in the PACKAGEID section:
PACKAGEID="Privoxy Team\Privoxy\Privoxy Package\X\Y\Z" |
You're now ready to build. Run:
os2build |
You will find the WarpIN-installable executable in the ./files directory. Upload this anonymously to uploads.sourceforge.net/incoming, create a release for it, and you're done. Use the release notes and Change Log from the source tarball package.
Login to Sourceforge's compilefarm via ssh:
ssh cf.sourceforge.net |
Choose the right operating system (not the Debian one). When logged in, make sure that you have freshly exported the right version into an empty directory. (See "Building and releasing packages" above). Then run:
cd current autoheader && autoconf && ./configure |
Then run
gmake solaris-dist |
which creates a gzip'ed tar archive. Sadly, you cannot use make solaris-upload on the Sourceforge machine (no ncftpput). You now have to manually upload the archive to Sourceforge's ftp server and release the file publicly. Use the release notes and Change Log from the source tarball package.
Note that the docbook generated files might need some hand editing, so the Windows build makefile does not rebuild the docs.
First, make sure that you have freshly exported the right version into an empty directory. (See "Building and releasing packages" above).
Then you can build the package. This is fully automated, and is controlled by windows/GNUmakefile. All you need to do is:
cd windows make |
Now you can manually rename privoxy_setup.exe to privoxy_setup_X.Y.Z.exe, and the build directory to privoxy_X.Y.Z. Create a .zip file of the newly renamed privoxy_X.Y.Z directory, GPG sign the installer and zip file,
$ gpg --armor --detach --sign privoxy_setup_X.Y.Z.exe $ gpg --armor --detach --sign privoxy_X.Y.Z.zip |
and upload the files to SourceForge.
When releasing the package on SourceForge, use the release notes and Change Log from the source tarball package.
First, make sure that you have freshly exported the right version into an empty directory. (See "Building and releasing packages" above). Then add a log entry to debian/changelog, if it is not already there, for example by running:
debchange -v 3.0.27-UNRELEASED-1 "New upstream version" |
Then, run:
dpkg-buildpackage -rfakeroot -us -uc -b |
This will create ../privoxy_3.0.27-UNRELEASED-1_i386.deb which can be uploaded. To upload the package to Sourceforge, simply issue
make debian-upload |
First, make sure that you have freshly exported the right version into an empty directory. (See "Building and releasing packages" above).
There are three modules available in the Git repository for use on Mac OS X, though technically only two of them generate a release (the other can be used to install from source).
The OSXPackageBuilder module generates OS X installer packages supporting all Macs running OS X 10.4 and above. Obtain it from Git as follows into a folder parallel to the exported privoxy source:
cvs -z3 -d:pserver:anonymous@ijbswa.cvs.sourceforge.net:/cvsroot/ijbswa co OSXPackageBuilder |
The module contains complete instructions on its usage in the file OS X Package Builder HOWTO.txt.
Once the package(s) have been generated, you can then upload them directly to the Files section of the Sourceforge project in the Macintosh (OS X) folder. Each new version release of Privoxy should have a new subfolder created in which to store its files. Please ensure that the folder contains a readme file that makes it clear which package is for whichversion of OS X.
This module is deprecated since the installer it generates places all Privoxy files in one folder in a non-standard location, and supports only Intel Macs running OS X 10.6 or higher.
Check out the module from Git as follows into a folder parallel to the exported privoxy source:
cvs -z3 -d:pserver:anonymous@ijbswa.cvs.sourceforge.net:/cvsroot/ijbswa co osxsetup |
Then run:
cd osxsetup build |
This will run autoheader, autoconf and configure as well as make. Finally, it will copy over the necessary files to the ./osxsetup/files directory for further processing by PackageMaker.
Bring up PackageMaker with the PrivoxyPackage.pmsp definition file, modify the package name to match the release, and hit the "Create package" button. If you specify ./Privoxy.pkg as the output package name, you can then create the distributable zip file with the command:
zip -r privoxyosx_setup_x.y.z.zip Privoxy.pkg |
You can then upload this file directly to the Files section of the Sourceforge project in the Macintosh (OS X) folder. Each new version release of Privoxy should have a new subfolder created in which to store its files. Please ensure that the folder contains a readme file that makes it clear which version(s) of OS X the package supports.
The macsetup module is ideal if you wish to build and install Privoxy from source on a single machine.
Check out the module from Git as follows into a folder parallel to the exported privoxy source:
cvs -z3 -d:pserver:anonymous@ijbswa.cvs.sourceforge.net:/cvsroot/ijbswa co macsetup |
The module contains complete instructions on its usage in its README file. The end result will be the exported version of Privoxy installed on the build machine.
Update the www/privoxy port and submit a diff upstream. For details see the FreeBSD Porter's Handbook.
After the package is ready, it is time to upload it to SourceForge, and go through the release steps. The upload is done via FTP:
Upload to: ftp://upload.sourceforge.net/incoming
user: anonymous
password: ijbswa-developers@lists.sourceforge.net
Or use the make targets as described above.
Once this done go to https://sourceforge.net/project/admin/editpackages.php?group_id=11118, making sure you are logged in. Find your target platform in the second column, and click Add Release. You will then need to create a new release for your package, using the format of $VERSION ($CODE_STATUS), e.g. 3.0.27 (beta).
Now just follow the prompts. Be sure to add any appropriate Release notes. You should see your freshly uploaded packages in "Step 2. Add Files To This Release". Check the appropriate box(es). Remember at each step to hit the "Refresh/Submit" buttons! You should now see your file(s) listed in Step 3. Fill out the forms with the appropriate information for your platform, being sure to hit "Update" for each file. If anyone is monitoring your platform, check the "email" box at the very bottom to notify them of the new package. This should do it!
If you have made errors, or need to make changes, you can go through essentially the same steps, but select Edit Release, instead of Add Release.
When all (or: most of the) packages have been uploaded and made available, send an email to the announce mailing list, Subject: "Version X.Y.Z available for download". Be sure to include the download location, the release notes and the Changelog. Also, post an updated News item on the project page Sourceforge, and update the Home page and docs linked from the Home page (see below). Other news sites and release oriented sites, such as Freshmeat, should also be notified.
Then update the source code for the next version to be released:
Increment the version number and change the code status to "UNRELEASED" in configure.in
Rebuild configure ("autoheader && autoconf") and GNUMakefile ("./configure")
"make dok-release" to update the sgml documentation source files.
Commit all your changes.