Software Configuration Management

If you want to use a software configuration management system to do cross-platform development, certain important considerations apply. This article, the first in a four-part series, takes an in-depth look at the requirements of an SCM system for cross-platform projects. It is excerpted from chapter three of the book Cross-Platform Development in C++, written by Syd Logan (Addison-Wesley; ISBN: 032124642X).

A software configuration management (SCM) system consists of tools and processes used to manage the source code of a project and to assess its state. With these tools, management and developers can do the following:

  1. Peruse the source code managed by the SCM system 
     
  2. Obtain a local copy of the source code 
     
  3. Submit modifications made to the source code by developers 
     
  4. Understand which changes have been made to the source code b y developers 
     
  5. Determine the state of the source code (for example, whether it compiles) 
     
  6. Log and track defects associated with binaries created from the source code

In a cross-platform project, certain aspects of an SCM system take on added importance. In this chapter, I describe the tools that were most important to Netscape and Mozilla in this regard:

  1. CVS, the Concurrent Version System, which was used to maintain and manage a central copy of the Netscape/Mozilla source code 
     
  2. Bugzilla, used to track defects in programs generated from the source code stored in CVS 
     
  3. Tinderbox, used to determine the state of the source code as changes were made to the copy stored in CVS 
     
  4. Patch, a tool that fosters the distribution and communication of changes among developers (in particular, between a developer and code reviewers) prior to it being permanently stored in CVS

These tools, combined with some best practices for using them that I describe in this chapter, helped greatly in ensuring that Netscape shipped cross-platform products that were similar in features and overall quality, and did so on the same schedule.

{mospagebreak title=Item 11: Use a Cross-Platform Bug Reporting and Tracking System}

A major component of a cross-platform toolset is the bug reporting and tracking system (which I refer to as a “bug system” here for brevity). A bug system is used by developers and testers to report defects and issues encountered during the software development and testing phases of a project. In general, a bug system should allow the reporter to specify the problem encountered, identify the context or state of the system at the time the bug was discovered, and list the steps required to reproduce the bug. The bug system also must allow for the tracking of any issues it contains. In terms of tracking, bugs generally go through the following states:

  1. New. A bug that has been discovered, but has not been investigated by a developer.
  2. Assigned. The developer acknowledges the bug and is investigating. 
     
  3. Resolved. The developer has fixed the bug, or has some other resolution (for example, unable to reproduce the problem, or the feature is working as designed). 
     
  4. Verified. QA/test has acknowledged the resolution of the developer, and has verified it to be correct.

In addition, a bug can be in the reopened state, an indication that a once-verified bug has resurfaced in testing after originally being verified by QA.

Finally, a bug system should allow someone to specify the relative priority of bugs filed against a product. Initially, this will be the person filing the bug, but ideally, it is done in a “triage” session, with participation of the following teams: development, QA/test, and product management. Tracking priorities is not easy, it takes discipline for people to sit down for an hour or two and grind through a list of bugs, but doing so allows developers to prioritize their work, and allows those who are making ship/no-ship decisions the opportunity to objectively decide what state the project is in.

When it comes to cross-platform development, you should always look for a couple of attributes and features when selecting a bug system for use in your project.

Accessibility

Perhaps the most important cross-platform attribute you should look for when selecting a bug system is accessibility. In a cross-platform development project, everyone will have his or her preferred development platform (see Item 1). For some developers, it will be Mac OS X, and for many it will be Windows. If the bug system is not accessible to everyone on the team via their native platforms, it risks not being used by those who would rather not boot into one of the operating systems that the bug system does support. Therefore, the bug system itself needs to be cross-platform. And, it must support all of your tier-1 platforms.

These days, by and large, the accessibility requirement is best met by using a Web-based bug system. Web browsers are available for every platform, including PDAs and mobile phones. Bug reporting/tracking systems based on LAMP (Linux, Apache, MySQL, PHP) or Java/JSP are available.

Ability to Track Platform-Specific Bugs

In some cases, a bug will be reproducible on all platforms, and in other cases, only a subset of the supported platforms will exhibit the given problem. It is critical that the platforms affected by the problem be tracked. The bug system should allow the person filing the bug, or even someone updating the bug at a later time, to specify the platform(s) affected by any given bug. Doing so will allow someone searching the bug database to specify a subset of the supported platforms as search criteria (for example, “find all assigned Linux bugs”). By specifying the platforms as part of the search criteria, platform specialists will be able to quickly identify bugs that affect their platform, and management will be able determine the number of open (and resolved) issues on a per-platform basis.

Bugzilla

Most of my experience with bug reporting systems has been with a bug tracking system named Bugzilla—yet another result of the Netscape/Mozilla project. Bugzilla is implemented as a Web application, allowing the database to be viewed from any platform that supports a Web browser. Bugzilla has been in active development for years, and is used by hundreds of organizations. Because Bugzilla originated during the development of the cross-platform Netscape/Mozilla system (it replaced a Netscape-internal bug system in use during the development of the 4.x browser called bugsplat), it has always supported cross-platform development directly.

Installing Bugzilla is pretty straightforward, and is well documented. A basic installation of Bugzilla can be done in about an hour, depending on your setup. (That was my experience on a stock Red Hat Fedora Core 4 system, and your results may vary.)

The following instructions are based on a Fedora Core 4 setup, using Apache and MySQL. (The default versions of Apache, Perl, and MySQL that come with FC4 are suitable for use with Bugzilla 2.22.)

Visit www.bugzilla.org, and download the latest stable release (2.22 as of this writing) onto the Linux machine that you will use to host the Bugzilla server. Then:

$ su
# tar -zxvf bugzilla-2.22.tar.gz
# mkdir /var/www/html/bugzilla
# cd bugzilla-2.22
# cp -r * /var/www/html/bugzilla
# cd /var/www/html/
# chown -R apache bugzilla
# chgrp -R apache bugzilla

The next step is to install required Perl modules. To do this, enter the following:

# perl -MCPAN -e ‘install “Bundle::Bugzilla"’

You may be prompted for inputs. If so, just take whatever default responses are offered. To validate that the required modules are installed, issue the following (as root):

# cd /var/www/html/bugzilla
# perl checksetup -check-modules

When that completes, you need to rerun checksetup once again:

# perl checksetup.pl

This will create a file named localconfig. To configure Bugzilla to work with MySQL, follow the instructions at www.bugzilla.org/docs/2.22/html/ configuration.html#mysql. Finally, rerun checksetup.pl one last time. The checksetup.pl script will create the Bugzilla database and prompt you for information needed to set up an administrator account in the Bugzilla database. Following this, you should be able to use the following URL to access bugzilla from your local Web browser:

  http://localhost/bugzilla

If you have any problems with these instructions (or have a setup that differs from mine), refer to file://var/www/html/bugzilla/docs/index.html for help.

After you have installed Bugzilla, you need to add products to the bug database, and modify the list of platforms (hardware and operating systems) that can be specified when bugs are filed in the database. To do this, log in as the administrator; you should see a screen similar to the one shown in Figure 3-1.


Figure 3-1.  Bugzilla main screen

In the bottom half of Figure 3-1, there is a link labeled Products. Clicking that link takes you to a screen on which you can enter the name and attributes of a product. Enter the product name and a description, accept defaults for the remaining fields, and click the Add button. You should get something like Figure 3-2.


Figure 3-2.  Products screen

The next step manages the list of supported operating systems. To view the list, click the Field Values link (visible in Figures 3-1 and 3-2). Then, click the OS link in the resulting page, and you should see the list of supported operating systems (All, Windows, Mac OS, Linux, and Other, as in Figure 3-3). By clicking the operating system names to edit them, change the value Windows to Windows XP, the value Mac OS to be Mac OS X 10.4, and the value Linux to be Fedora Core 4. I personally think it is better to be fine-grained when it comes to the list of operating systems; Linux, for example, is far too vague to be truly useful. You can also add or remove operating systems using the provided controls.


Figure 3-3.   Platforms screen

To modify the list of supported hardware, click the Field Values link again, click the Hardware link on the resulting page, and then modify the list of supported hardware types so that it is consistent with your supported hardware platforms. Again, being fine-grained can help in diagnosing problems. For example, instead of Macintosh, you might add choices based on type and CPU (for example, PowerBook and MacBook).

Before any bug can be filed against a product, components and version numbers need to be added for the product. This can be done by clicking the Product link, clicking the product name in the resulting table, and then clicking the Edit Components and Edit Versions links, respectively. Components are specific areas of a product. For Trixul, I defined three components: JavaScript Integration, Layout engine, and liblayout. Additional components can be added by the administrator as needed. Version numbers can be any text string that helps to identify the versions of the product; I specified 0.1.0 and 1.0. Figure 3-4 illustrates the resulting product summary page displayed by Bugzilla. As soon as at least one component and one version number has been specified, users should be able to add new bugs to the database.


Figure 3-4.  Product Summary Page

Returning to the cross-platform issues of platform and hardware, let’s see how they are specified by filing a bug against Trixul. One problem with Trixul is that under Microsoft Windows XP, the layout engine is built as an executable, whereas on all other platforms, the layout engine is built as a shared library that is then linked to a small executable. The reason for this bug (I believe) is a difficulty with linking a mixed (native and Common Language Infrastructure [CLI]) .NET application to a shared library using Visual Studio .NET 2003.

To file this bug, I logged on to Bugzilla, clicked the Enter a New Bug Report link, filled out the form as shown in Figure 3-5, and then clicked Commit.

Notice how I selected Windows XP in the OS field, and PC (Intel Pentium 4) in the Platform field.

Platform-specific bugs can be easily located using the search feature of Bugzilla. To do so, visit the main Bugzilla page and click the Search Existing Bug Reports link. The resulting page allows you to specify a wide variety of search criteria, including the OS and Platform fields. Figure 3-6 illustrates a search of all open bugs filed against both Fedora Core 4 and Windows XP on the Pentium 4 platform.


Figure 3-5.   Filing a new bug report


Figure 3-6.  Searching for bugs

If you like, you can copy and paste the URL of the search result page into a Web page or e-mail, and use that as a quick way to find all bugs related to a specific platform/operating system combination. When using Bugzilla I generally make a Web page consisting of links to all queries that I consider essential for the project I am working on; clicking a link is much easier than revisiting the Bugzilla search page and reentering all the parameters of the desired query each time I go to look at my list of bugs.

That’s pretty much it for Bugzilla. Using the general guidelines I just provided, it is fairly easy to set up and maintain a database that is capable of helping you ensure that platform-specific issues are tracked accurately.

Please check back for the next part of the series.

[gp-comments width="770" linklove="off" ]

antalya escort bayan antalya escort bayan Antalya escort diyarbakir escort