Running Boost Regression Tests

This runs all regressions and reports the results back to the Boost community.

Requirements

• Python 2.3 or later.
• Subversion 1.4 or later (optional).
• At least 5 gigabytes of disk space per compiler to be tested.

Step by step instructions

1. Create a new directory for the branch you want to test.
2. Download the run.py script into that directory.

3. Run "python run.py [options] [commands]" with at minimum the two options:

   • --runner - Your choice of name that identifies your results in the reports ^{1, 2}.
   • --toolsets - The toolset(s) you want to test with ³.

   For example:

   python run.py --runner=Metacomm --toolsets=gcc-4.2.1,msvc-8.0

Note: If you are behind a firewall/proxy server, everything should still "just work". In the rare cases when it doesn't, you can explicitly specify the proxy server parameters through the --proxy option, e.g.:

python run.py ... --proxy=http://www.someproxy.com:3128

Options

commands: cleanup, collect-logs, get-source, get-tools, patch, regression,
setup, show-revision, test, test-boost-build, test-clean, test-process, test-
run, update-source, upload-logs

Options:
  -h, --help            show this help message and exit
  --runner=RUNNER       runner ID (e.g. 'Metacomm')
  --comment=COMMENT     an HTML comment file to be inserted in the reports
  --tag=TAG             the tag for the results
  --toolsets=TOOLSETS   comma-separated list of toolsets to test with
  --libraries=LIBRARIES
                        comma separated list of libraries to test
  --incremental         do incremental run (do not remove previous binaries)
  --timeout=TIMEOUT     specifies the timeout, in minutes, for a single test
                        run/compilation
  --bjam-options=BJAM_OPTIONS
                        options to pass to the regression test
  --bjam-toolset=BJAM_TOOLSET
                        bootstrap toolset for 'bjam' executable
  --pjl-toolset=PJL_TOOLSET
                        bootstrap toolset for 'process_jam_log' executable
  --platform=PLATFORM   
  --user=USER           Boost SVN user ID
  --local=LOCAL         the name of the boost tarball
  --force-update        do an SVN update (if applicable) instead of a clean
                        checkout, even when performing a full run
  --have-source         do neither a tarball download nor an SVN update; used
                        primarily for testing script changes
  --ftp=FTP             FTP URL to upload results to.
  --proxy=PROXY         HTTP proxy server address and port
                        (e.g.'http://www.someproxy.com:3128')
  --ftp-proxy=FTP_PROXY
                        FTP proxy server (e.g. 'ftpproxy')
  --dart-server=DART_SERVER
                        the dart server to send results to
  --debug-level=DEBUG_LEVEL
                        debugging level; controls the amount of debugging
                        output printed
  --send-bjam-log       send full bjam log of the regression run
  --mail=MAIL           email address to send run notification to
  --smtp-login=SMTP_LOGIN
                        STMP server address/login information, in the
                        following form: <user>:<password>@<host>[:<port>]
  --skip-tests          do not run bjam; used for testing script changes

To test trunk use "--tag=trunk" (the default), and to test the release use "--tag=branches/release". Or substitute any Boost tree of your choice.

Details

The regression run procedure will:

• Download the most recent regression scripts.
• Download the designated testing tool sources including Boost.Jam, Boost.Build, and the various regression programs.
• Download the most recent from the Boost Subversion Repository into the subdirectory boost.
• Build bjam and process_jam_log if needed. (process_jam_log is a utility, which extracts the test results from the log file produced by Boost.Build).
• Run regression tests, process and collect the results.
• Upload the results to a common FTP server.

The report merger process running continuously will merge all submitted test runs and publish them at various locations.

Advanced use

Providing detailed information about your environment

Once you have your regression results displayed in the Boost-wide reports, you may consider providing a bit more information about yourself and your test environment. This additional information will be presented in the reports on a page associated with your runner ID.

By default, the page's content is just a single line coming from the comment.html file in your run.py directory, specifying the tested platform. You can put online a more detailed description of your environment, such as your hardware configuration, compiler builds, and test schedule, by simply altering the file's content. Also, please consider providing your name and email address for cases where Boost developers have questions specific to your particular set of results.

Incremental runs

You can run run.py in incremental mode by simply passing it an identically named command-line flag:

python run.py ... --incremental

Getting sources from Tarball

By default the sources are obtained from the Boost Subversion Repository and we prefer testers use SVN. But if you can't have an SVN client installed you can obtain the sources as tarballs (*.tar.gz). To indicate this pass an empty user to run.py:

python run.py ... --user=

Note: Both methods to obtain the sources will get the latest code. This is accomplished by building the tarball on-demand from the SVN sources.

Patching Boost sources

You might encounter an occasional need to make local modifications to the Boost codebase before running the tests, without disturbing the automatic nature of the regression process. To implement this under regression.py:

1. Codify applying the desired modifications to the sources located in the ./boost subdirectory in a single executable script named patch_boost (patch_boost.bat on Windows).
2. Place the script in the run.py directory.

The driver will check for the existence of the patch_boost script, and, if found, execute it after obtaining the Boost sources.

Feedback

Please send all comments/suggestions regarding this document and the testing procedure itself to the Boost Testing list.

Notes

[1] If you are running regressions interlacingly with a different set of compilers (e.g. for Intel in the morning and GCC at the end of the day), you need to provide a different runner id for each of these runs, e.g. your_name-intel, and your_name-gcc.

[2] The limitations of the reports' format/medium impose a direct dependency between the number of compilers you are testing with and the amount of space available for your runner id. If you are running regressions for a single compiler, please make sure to choose a short enough id that does not significantly disturb the reports' layout. You can also use spaces in the runner ID to allow the reports to wrap the name to fit.

[3] If --toolsets option is not provided, the script will try to use the platform's default toolset (gcc for most Unix-based systems).

[4] By default, the script runs in what is known as full mode: on each run.py invocation all the files that were left in place by the previous run — including the binaries for the successfully built tests and libraries — are deleted, and everything is rebuilt once again from scratch. By contrast, in incremental mode the already existing binaries are left intact, and only the tests and libraries which source files has changed since the previous run are re-built and re-tested.

The main advantage of incremental runs is a significantly shorter turnaround time, but unfortunately they don't always produce reliable results. Some type of changes to the codebase (changes to the bjam testing subsystem in particular) often require switching to a full mode for one cycle in order to produce trustworthy reports.

As a general guideline, if you can afford it, testing in full mode is preferable.