Difference between revisions of "Information for Developers"

From PKP Wiki
Jump to: navigation, search
m (Corrected a typo.)
(PKP library submodule changes: Clarify the doc)
 
(11 intermediate revisions by 6 users not shown)
Line 1: Line 1:
= Bugzilla =
+
= Managing Contributions =
 +
 
 +
PKP code development, including external contributions, is managed between [http://pkp.sfu.ca/bugzilla Bugzilla] and [https://github.com/pkp/ Github]. We prefer that issues be opened as either bugs or feature requests in Bugzilla, and that fixes are submitted as pull requests via Github, but we do still accept patches uploaded to Bugzilla itself if you do not want to submit via Github pull request.
 +
 
 +
= Identifying an Issue =
  
 
You can access our Bugzilla database [http://pkp.sfu.ca/bugzilla here]. We welcome any bug reports and feature requests, so long as they are understandably written and flagged correctly (mostly an issue of setting the severity between enhancement; trivial; minor; normal; major; critical; or blocker levels, although you can also set the priority as well if you'd like). You can see a great set of bug-writing guidelines [http://pkp.sfu.ca/bugzilla/page.cgi?id=bug-writing.html here].  
 
You can access our Bugzilla database [http://pkp.sfu.ca/bugzilla here]. We welcome any bug reports and feature requests, so long as they are understandably written and flagged correctly (mostly an issue of setting the severity between enhancement; trivial; minor; normal; major; critical; or blocker levels, although you can also set the priority as well if you'd like). You can see a great set of bug-writing guidelines [http://pkp.sfu.ca/bugzilla/page.cgi?id=bug-writing.html here].  
Line 5: Line 9:
 
You are also advised to [http://pkp.sfu.ca/bugzilla/query.cgi search] for similar reports to avoid duplication.  
 
You are also advised to [http://pkp.sfu.ca/bugzilla/query.cgi search] for similar reports to avoid duplication.  
  
= CVS Access =
+
= Contributing Code =
  
== Web-based CVS ==  
+
== Contributing via Github ==
  
You can browse our CVS repository online [http://pkp.sfu.ca/cvs/cvsweb.cgi/#dirlist here].  
+
Contributing code via Github is our '''preferred method'''. Please see http://pkp.sfu.ca/wiki/index.php/HOW-TO_check_out_PKP_applications_from_git and http://pkp.sfu.ca/wiki/index.php/Frequent_git_use_cases for extensive instructions on the ins and outs of working with git repositories.
  
== Command Line CVS access ==
+
To contribute code back to us, feel free to submit a pull request from git.
  
(This is not a guide on how to use CVS. For more CVS information, try [http://www.nongnu.org/cvs/ here], [http://www.eyrie.org/~eagle/notes/cvs/basic-usage.html here], and [http://en.wikipedia.org/wiki/Concurrent_Versions_System here].
+
If you are working against a particular entry in our Bugzilla database, use the following format for commit messages:
  
You can check out all CVS modules via anonymous CVS. Before you do so, you should know which modules, exactly, you want to work with. OJS and OCS each have a stable and devel branch; the stable branches are standalone maintenance releases (extensions of OJS 2.2.2 and OCS 2.1.1 codebases), while the devel branches also requires the PKP Web Application Library (WAL) be checked out and installed in the lib/ directory. Harvester2 and Open Monograph Press only have one devel branch, and each require the PKP WAL. Lemon8 is a standalone program with only one branch at the moment.
+
    *1234* My Commit Description
  
=== Setting up the Environment ===
+
...where the 1234 is the bug ID from Bugzilla. This allows Bugzilla to track back to the commits made in that entry and helps us assemble release notes consistently.
  
The following instructions are more or less directly applicable to any *nix operating systems (OS X incl.); Windows users probably need to do things a little differently. Also, these are general instructions, pertaining to how I manage my own setup -- you might want to do things a little bit differently.
+
=== PKP library submodule changes ===
  
Firstly, create a CVS directory to store all your cvs checkouts in. I created mine in /Users/jmacgreg/ and then entered into the new directory, from which I ran the remaining commands:
+
If you changed something in library, you will also need to commit the library submodule changes in the application repository. This is necessary especially for our continuous integration structure. Basically, the submodule changes commit will tell anyone that pulls the application code where the submodule repository HEAD is, including the CI script.
  
mkdir /Users/jmacgreg/cvs
+
This is needed because the CI script is prepared to get the library from PKP's official repository. In case of a pull request from a developer, we need to tell the CI script to also update the library with commits from the developer's library repository. That way the CI tool can correctly build and test your pull request code before merging. To make that happen, do the following:
cd cvs
+
  
You'll then have to log into our CVS repository using our anonymous credentials:
+
* create your commits normally, both for application and library;
 +
* if you have library commits, commit the submodule changes, at the application repository, in an UNIQUE separate commit, making sure that it is the HEAD:
 +
<pre>
 +
git add lib/pkp
 +
git commit -m '*1234* Any message here ##githubUser/branch##'
 +
</pre>
 +
1234 is the bug ID;
 +
githubUser is your github user;
 +
branch is the branch in your github repository where the library commits are;
 +
* make sure your local application and library repositories are updated with the official, as it is expected if you want to do a pull request;
 +
* push both library and application commits to your repository in github;
  
cvs -d :pserver:anonymous@lib-pkp.lib.sfu.ca:/cvs login
+
After doing this, you can create a pull request using the GitHub interface, and check the CI build result. If it doesn't pass, you can read the logs, see the errors, fix the code and update your pull request by pushing your code to your own repository again.
  
You will be asked for a password — there is none, so just hit enter. After which you should be able to run the following commands to grab specific modules.
+
== Contributing Patches ==
  
=== OJS and OCS '''stable''' branches ===
+
If you cannot or do not want to contribute via Github, you can upload patches directly to bug reports. We're always happy to receive patches, whether for contribution to a project or for general community availability on [http://pkp.sfu.ca/support/forum the forums]. The one thing we do ask is that you provide the patch in [http://en.wikipedia.org/wiki/Diff#Unified_format Unified Diff] format. If you are using our [http://pkp.sfu.ca/wiki/index.php/Main_Page#Development_Topics git instructions] to work with the application codebase (highly recommended), you can create unified diffs via git using the following command:
  
To grab the OCS and OJS stable branches, run the following commands:
+
git diff -u ./ > patch.diff
  
cvs -d :pserver:anonymous@lib-pkp.lib.sfu.ca:/cvs checkout -d ojs2-stable -r ojs2-branch-2_2_2 ojs2
+
... which will find all differences in the working directory and '''recursively''', and write them to a file called patch.diff.
  
The above command will download the current code for the OJS 2.2.2 branch of the ojs2 module, and deposit in a new 'ojs2-stable' directory in the directory you are currently in (you moved to your cvs/ directory, right?)
 
  
cvs -d :pserver:anonymous@lib-pkp.lib.sfu.ca:/cvs checkout -d ocs2-stable -r ocs2-branch-2_1_1 ocs2
+
== Applying Patches ==
  
The above command will download the OCS 2.1.1 code into ocs2-stable/.
+
=== Patching Manually ===
  
Setting this up to be served by Apache is up to you (I have it symlinked to my web folder). Remember, while this branch is still fairly stable in comparison to the heavily modified devel branch, it's still a work in progress and shouldn't be used in a production environment unless you know what you're doing.
+
If you want to apply a patch manually, whether from Bugzilla or from Github, download it to your development environment, and from the application's root directory and run
  
=== Lemon8-XML ===
+
patch -p0 < ./path/to/patch.diff
  
To checkout the most recent Lemon8 code, run
+
The --dry-run option to the patch tool allows you to safely test the patch to see if it will apply cleanly, and if not, where the conflicts will arise.
  
cvs -d :pserver:anonymous@lib-pkp.lib.sfu.ca:/cvs checkout -d lemon8-xml lemon8-xml
+
For more Windows (and general) information, also see [http://pkp.sfu.ca/support/forum/viewtopic.php?f=8&t=3779 this forum thread].
  
The above command will download the latest Lemon8 development code into a new lemon8-xml directory.
+
=== Patching from Github ===
  
=== Development Branches ===
+
You can grab "patches" from Github by finding that particular commit's has and cherry-picking it:
  
OJS and OCS development branches, as well as the only Harvester2 and OMP (and de facto development) branches, all work a little differently: they all rely on the PKP WAL module as a dependency. The WAL module needs to be 'installed' (symlinking is fine) in each branches' lib/ directory.  
+
First, find the hash for the commit you want to apply locally. The hash is the last string in the commit's URL, for example for the commit found at https://github.com/pkp/ojs/commit/4aba081d35133c1ee9a9e1f3b9166ca5db5fd48d it would be "4aba081d35133c1ee9a9e1f3b9166ca5db5fd48d".  
  
First, checkout the packages you need:  
+
Next, make sure your local repository is up to date (NB that this does not automatically merge all new remote commits to your repository, it [http://stackoverflow.com/questions/292357/whats-the-difference-between-git-pull-and-git-fetch just stores them locally]):  
  
'''OJS devel''' (2.3):
+
git fetch
  
cvs -d :pserver:anonymous@lib-pkp.lib.sfu.ca:/cvs checkout -d ojs2-devel ojs2
+
Then, cherry-pick the commit you want to apply by hash:  
  
 +
git cherry-pick 4aba081d35133c1ee9a9e1f3b9166ca5db5fd48d
  
'''Harvester devel''' (2.3):
+
= Setting up a flexible development environment =
 
+
cvs -d :pserver:anonymous@lib-pkp.lib.sfu.ca:/cvs checkout -d harvester2 harvester2
+
 
+
 
+
'''OMP devel''' (2.3):
+
 
+
cvs -d :pserver:anonymous@lib-pkp.lib.sfu.ca:/cvs checkout -d omp omp
+
 
+
 
+
Now the tricky part. You need to download the PKP WAL (called 'pkp' in CVS) and either copy it or [http://en.wikipedia.org/wiki/Symbolic_link symlink] it to your devel instances' lib/ directory. I prefer symlinking, as I can do this for each devel instance I'm running, and only have to remember to update one directory.
+
 
+
Checkout the pkp module:
+
 
+
cvs -d :pserver:anonymous@lib-pkp.lib.sfu.ca:/cvs checkout -d pkp pkp
+
 
+
Symlink the module to eg. OJS' lib/ directory:
+
 
+
ln -s /Users/jmacgreg/cvs/pkp /Users/jmacgreg/cvs/ojs2-devel/lib/pkp
+
 
+
... and repeat as necessary for each application.
+
 
+
Alternatively, it is possible to change into the ojs2-devel/lib and create, what are called, relative symlinks:
+
 
+
  cd ojs2-devel/lib
+
  ln -s ../../pkp
+
 
+
= Patches =
+
 
+
== Making Patches ==
+
 
+
We're always happy to receive patches, whether for contribution to a project or for general community availability on [http://pkp.sfu.ca/support/forum the forums]. The one thing we do ask is that you provide the patch in [http://en.wikipedia.org/wiki/Diff#Unified_format Unified Diff] format. If you are patching against one of the above CVS modules, the easiest thing to do is generate your patch using something like the following command (which will generate a diff of the entire directory you are currently in, for example ojs2-devel, against the entire cvs module):
+
 
+
cvs diff -uN ./ > patch.diff
+
 
+
=== Applying Patches ===
+
 
+
If you want to apply a patch, download it to your development environment, and from the application's root directory and run
+
 
+
patch -p0 < ./path/to/patch.diff
+
 
+
The --dry-run option to the patch tool allows you to safely test the patch to see if it will apply cleanly, and if not, where the conflicts will arise.
+
 
+
For more Windows (and general) information, also see [http://pkp.sfu.ca/support/forum/viewtopic.php?f=8&t=3779 this forum thread].
+
 
+
= Testing your code with all supported environments =
+
  
 
NB: This is not a beginner's installation/configuration tutorial. We only highlight a few important differences from a standard installation as a help for experienced PKP, database, PHP and web server users.
 
NB: This is not a beginner's installation/configuration tutorial. We only highlight a few important differences from a standard installation as a help for experienced PKP, database, PHP and web server users.
Line 135: Line 103:
 
** extension_dir must point to the absolute path of the version specific extension binaries
 
** extension_dir must point to the absolute path of the version specific extension binaries
 
** enable extensions required by PKP software (e.g. database client, GD, etc.)
 
** enable extensions required by PKP software (e.g. database client, GD, etc.)
** for PHP 4's php.ini only: cgi.fix_pathinfo=1
 
 
** optional: error_reporting, display_errors, log_errors, error_log, max_execution_time, xdebug
 
** optional: error_reporting, display_errors, log_errors, error_log, max_execution_time, xdebug
 
* Test all PHP installations separately: /path/to/php -c /path/to/php.ini -v (This should discover most problems with wrong extension paths).
 
* Test all PHP installations separately: /path/to/php -c /path/to/php.ini -v (This should discover most problems with wrong extension paths).
Line 237: Line 204:
 
  http -DPHP52 -DMODULE
 
  http -DPHP52 -DMODULE
 
  http -DPHP53 -DFCGI
 
  http -DPHP53 -DFCGI
 
Please note that PHP 4.3 only supports CGI mode. PHP 4's module is only compatible with Apache 1.3 or 2.0 and PHP 4 does not correctly interpret fcgi's PATH_INFO when used with the FcgiWrapper directive.
 
 
= PHP 4 compatibility coding oddities collection =
 
 
We'll not try to reproduce a full catalogue of PHP4/5 incompatibilities here. Have a look at http://www.php.net/manual/en/migration5.php which is quite complete. There are however some PKP-specific solutions to certain compatibility problems that we document here.
 
 
== Chaining -> accessors ==
 
 
PHP 4 does not support the chaining of -> accessors. We therefore write:
 
 
$someTempVar =& $firstObject->getOtherObject();
 
$someTargetVar =& $someTempVar->getTargetObject();
 
 
== Passing objects as method parameters ==
 
 
Most of the more serious PHP4 compatibility problems have to do with the fact that PHP5 passes and assigns objects by reference by default while PHP4 makes shallow copies if not explicitly asked to use references. This results in a few coding particularities that I'll describe here.
 
 
We always use explicit by-ref method parameters when possible. This will automatically guarantee PHP4/5 compatibility:
 
 
function myMethod(&myObject) { ... }
 
 
== Assigning objects to variables by-ref ==
 
 
We always assign objects explicitly by-ref when possible. This will automatically guarantee PHP4/5 compatibility:
 
 
$myObject =& $myOtherObject;
 
 
== Default values for by-ref  method parameters ==
 
 
PHP 5 allows default values for by-ref method parameters while PHP 4 doesn't. Default parameters for by-ref objects usually don't make sense. There's one notable exception though: Initializing a variable with null when the by-ref parameter is optional. As PHP4 doesn't support assinging default values to default parameters, we'll write
 
 
function myMethod(myObject = null) {...}
 
 
to maintain PHP4 compatibility although this is less efficient in PHP4.
 
 
== Returning parameters by-ref ==
 
 
Returning non-variable return values by-ref in PHP 5 generates a warning while PHP 4 tolerates this. We therefore always use the following workaround:
 
 
function &myMethod() {
 
    ...
 
    $returner = null;
 
    return $returner;
 
}
 
 
== Use of $this in the constructor ==
 
 
This is a tricky one. To understand it look at the following code:
 
 
class Test {
 
    var $instance;
 
    var $data;
 
 
    function Test() {
 
        $this->instance =& $this;
 
    }
 
}
 
 
$instanceFromNew = new Test();
 
$instanceFromConstructor =& $instanceFromNew->instance;
 
 
$instanceFromNew->data = 1;
 
$instanceFromConstructor->data = 2;
 
 
echo $instanceFromNew->data."\n";
 
echo $instanceFromConstructor->data;
 
 
This will result in the output 1 - 2 for PHP 4 and 2 - 2 for PHP 5.
 
 
Changing
 
$instanceFromNew = new Test();
 
to
 
$instanceFromNew =& new Test();
 
 
Solves the problem in PHP 4 but causes a deprecation warning in PHP 5.3.
 
 
This is especially problematic in our case as we use lambda functions for validation that are created in the constructor and often get $this passed in as a parameter. If we don't use ... =& new ... in the caller then changes made by the instantiating caller to the object later on will not appear to the validator in PHP 4. This leads to validation errors and can cause security vulnerabilities. In PHP 5.3 the deprecation warning will be raised on parse time. This means that the workaround no longer works. We have to switch off E_DEPRECATION in PHP 5.3 until we drop PHP 4 compatibility.
 
 
Our solution to this problem is:
 
 
* When you create a reference to $this within the constructor that will later be used to '''access data in the object instance that might be changed by the instantiating caller''' then you'll have to use the following workaround:
 
  if (checkPhpVersion('5.0.0')) { // WARNING: This form needs $this in constructor
 
      ... = new ...;
 
  } else {
 
      ... =& new ...;
 
  }
 
 
* In all other cases just use
 
... = new ...
 
in the caller. This is a little less efficient in PHP 4 but makes the code more readable and easier to maintain.
 
 
= git Access (Incubation) =
 
 
We are currently evaluating switching part of our internal development workflow to [http://git-scm.com/ git]. This is still work in progress. Only parts of Open Monograph Press (OMP) are currently developed in git.
 
 
We use git sub-modules to integrate our PKP application library into end-user applications. As this is a less known and a little tricky feature of git we developed a PKP-specific [[git sub-module tutorial]] for developers. Please follow the steps in this tutorial before starting to develop with git. It will help you to understand the basic concepts of git sub-modules and will also give you an idea of the general git repository set-up we use at PKP.
 
 
Once you have worked through the sub-module tutorial you may continue with [[HOW-TO check out PKP applications from git]]. This shows you the specific steps you should follow to set up a local PKP application development repository.
 
 
Those who love CVS but want to jump on git anyway should check out [[HOW-TO use git and CVS in parallel]].
 

Latest revision as of 15:30, 25 July 2014

Managing Contributions

PKP code development, including external contributions, is managed between Bugzilla and Github. We prefer that issues be opened as either bugs or feature requests in Bugzilla, and that fixes are submitted as pull requests via Github, but we do still accept patches uploaded to Bugzilla itself if you do not want to submit via Github pull request.

Identifying an Issue

You can access our Bugzilla database here. We welcome any bug reports and feature requests, so long as they are understandably written and flagged correctly (mostly an issue of setting the severity between enhancement; trivial; minor; normal; major; critical; or blocker levels, although you can also set the priority as well if you'd like). You can see a great set of bug-writing guidelines here.

You are also advised to search for similar reports to avoid duplication.

Contributing Code

Contributing via Github

Contributing code via Github is our preferred method. Please see http://pkp.sfu.ca/wiki/index.php/HOW-TO_check_out_PKP_applications_from_git and http://pkp.sfu.ca/wiki/index.php/Frequent_git_use_cases for extensive instructions on the ins and outs of working with git repositories.

To contribute code back to us, feel free to submit a pull request from git.

If you are working against a particular entry in our Bugzilla database, use the following format for commit messages:

   *1234* My Commit Description

...where the 1234 is the bug ID from Bugzilla. This allows Bugzilla to track back to the commits made in that entry and helps us assemble release notes consistently.

PKP library submodule changes

If you changed something in library, you will also need to commit the library submodule changes in the application repository. This is necessary especially for our continuous integration structure. Basically, the submodule changes commit will tell anyone that pulls the application code where the submodule repository HEAD is, including the CI script.

This is needed because the CI script is prepared to get the library from PKP's official repository. In case of a pull request from a developer, we need to tell the CI script to also update the library with commits from the developer's library repository. That way the CI tool can correctly build and test your pull request code before merging. To make that happen, do the following:

  • create your commits normally, both for application and library;
  • if you have library commits, commit the submodule changes, at the application repository, in an UNIQUE separate commit, making sure that it is the HEAD:
git add lib/pkp
git commit -m '*1234* Any message here ##githubUser/branch##'

1234 is the bug ID; githubUser is your github user; branch is the branch in your github repository where the library commits are;

  • make sure your local application and library repositories are updated with the official, as it is expected if you want to do a pull request;
  • push both library and application commits to your repository in github;

After doing this, you can create a pull request using the GitHub interface, and check the CI build result. If it doesn't pass, you can read the logs, see the errors, fix the code and update your pull request by pushing your code to your own repository again.

Contributing Patches

If you cannot or do not want to contribute via Github, you can upload patches directly to bug reports. We're always happy to receive patches, whether for contribution to a project or for general community availability on the forums. The one thing we do ask is that you provide the patch in Unified Diff format. If you are using our git instructions to work with the application codebase (highly recommended), you can create unified diffs via git using the following command:

git diff -u ./ > patch.diff

... which will find all differences in the working directory and recursively, and write them to a file called patch.diff.


Applying Patches

Patching Manually

If you want to apply a patch manually, whether from Bugzilla or from Github, download it to your development environment, and from the application's root directory and run

patch -p0 < ./path/to/patch.diff

The --dry-run option to the patch tool allows you to safely test the patch to see if it will apply cleanly, and if not, where the conflicts will arise.

For more Windows (and general) information, also see this forum thread.

Patching from Github

You can grab "patches" from Github by finding that particular commit's has and cherry-picking it:

First, find the hash for the commit you want to apply locally. The hash is the last string in the commit's URL, for example for the commit found at https://github.com/pkp/ojs/commit/4aba081d35133c1ee9a9e1f3b9166ca5db5fd48d it would be "4aba081d35133c1ee9a9e1f3b9166ca5db5fd48d".

Next, make sure your local repository is up to date (NB that this does not automatically merge all new remote commits to your repository, it just stores them locally):

git fetch 

Then, cherry-pick the commit you want to apply by hash:

git cherry-pick 4aba081d35133c1ee9a9e1f3b9166ca5db5fd48d

Setting up a flexible development environment

NB: This is not a beginner's installation/configuration tutorial. We only highlight a few important differences from a standard installation as a help for experienced PKP, database, PHP and web server users.

Install and Configure Databases

  • There is nothing special to the database installation. Use your OS' standard installation procedure to install the necessary databases you want to test in parallel.
  • Edit config.inc.php to switch the database connection.

Install and Configure supported PHP versions

  • Download or build the PHP binaries you want to test (newer versions can be downloaded from php.net or are part of the standard OS distributions, older versions must be built from source).
  • You'll need the thread-safe version if you want to test mod_php in Apache.
  • Install all binaries and configuration separately. Do not create any shared files or configuration (separate php.ini, separate installation directories).
  • On Windows:
    • Get the PHP zip arquive distributions where available and unzip them into separate folders.
    • XAMPP has lots of old Windows binaries if the version you look for is no longer officially supported on php.net.
    • Download the thread-safe VC6 version as VC9 versions are not compatible with Apache.
    • Do not install DLL files in C:\WINDOWS or any other directory on the path.
    • Do not install any registry keys.
    • Do not configure any shared environment variables.
  • Exception: Install PEAR in a central folder that is accessible by all PHP versions. Install necessary PEAR dependencies (e.g. PHPUnit, etc.) - see separate documentation on this Wiki for that.
  • Edit the development php.ini files as required:
    • extension_dir must point to the absolute path of the version specific extension binaries
    • enable extensions required by PKP software (e.g. database client, GD, etc.)
    • optional: error_reporting, display_errors, log_errors, error_log, max_execution_time, xdebug
  • Test all PHP installations separately: /path/to/php -c /path/to/php.ini -v (This should discover most problems with wrong extension paths).

Install and Configure Apache

We use Apache so that we can test CGI, FCGI and Module configurations in parallel. Install Apache 2.2 (standard distribution for your OS). Then get and install mod_fcgid as explained on the linked web pages.

All the magic lies in Apache's httpd.conf. Here's the Windows version. It should be easy to adapt that to your specific OS:

...
#General CGI Support
<IfDefine CGI>
    LoadModule cgi_module modules/mod_cgi.so
    AddHandler application/x-httpd-php .php
</IfDefine>
#General FCGI Support
<IfDefine FCGI>
    LoadModule fcgid_module modules/mod_fcgid.so
    AddHandler fcgid-script .php
    <Directory "C:/path/to/your/pkp/doc/root">
        Options +ExecCGI
    </Directory>
</IfDefine>
#PHP 4.3 support
<IfDefine PHP43>
    <IfDefine CGI>
        Action application/x-httpd-php "/php-cgi/php.exe"
        ScriptAlias /php-cgi/ "C:/path/to/php43/"
        <Directory "C:/path/to/php43">
            Order allow,deny
            Allow from 127.0.0.1
        </Directory>
    </IfDefine>
</IfDefine>
#PHP 5.2 support
<IfDefine PHP52>
    <IfDefine MODULE>
        LoadFile "C:/path/to/php52/php5ts.dll"
        LoadFile "C:/path/to/php52/libmysql.dll" # This is required (PHP52 only) otherwise PHP might load libmysql.dll from MySQL's own bin
        LoadModule php5_module "C:/path/to/php52/php5apache2_2.dll"
        PHPIniDir "C:/path/to/php52"
    </IfDefine>
    <IfDefine FCGI>
        FcgidWrapper "C:/path/to/php52/php-cgi.exe" .php
    </IfDefine>
    <IfDefine CGI>
        Action application/x-httpd-php "/php-cgi/php-cgi.exe"
        ScriptAlias /php-cgi/ "C:/path/to/php52/"
        <Directory "C:/path/to/php52">
            Order allow,deny
            Allow from 127.0.0.1
        </Directory>
    </IfDefine>
</IfDefine>
#PHP 5.3 support
<IfDefine PHP53>
    <IfDefine MODULE>
        LoadFile "C:/path/to/php53/php5ts.dll"
        LoadModule php5_module "C:/path/to/php53/php5apache2_2.dll"
        PHPIniDir "C:/path/to/php53"
    </IfDefine>
    <IfDefine FCGI>
        FcgidWrapper "C:/path/to/php53/php-cgi.exe" .php
    </IfDefine>
    <IfDefine CGI>
        Action application/x-httpd-php "/php-cgi/php-cgi.exe"
        ScriptAlias /php-cgi/ "C:/path/to/php53/"
        <Directory "C:/path/to/php53">
            Order allow,deny
            Allow from 127.0.0.1
        </Directory>
    </IfDefine>
</IfDefine>
DocumentRoot "C:/path/to/your/pkp/doc/root"
<Directory "C:/path/to/your/pkp/doc/root">
    Options +FollowSymLinks
    AllowOverride All
    Order allow,deny
    Allow from all
</Directory>
...

Now all you need to switch configurations is starting Apache like this:

http -D(PHP43|PHP52|PHP53) -D(MODULE|CGI|FCGI)

A few examples:

http -DPHP43 -DCGI
http -DPHP52 -DMODULE
http -DPHP53 -DFCGI