This repository contain a regression test suite for AbiWord.

Running the fully automated testsuite for AbiWord is as simple as calling

	./regression.pl

The results of the test are displayed on STDOUT, and can be in plain text format or HTML, depending on the value of the $html variable located near the top of the regression.conf configuration file.

The rest of this file describes in some more detail which tests are performed.

The raw diff test is a simple test to see if a file that is imported into abiword is converted into the same internal representation every time. Changes that are made to an importer can result in a different awml document for example. Some of the detected changes may be desired, others can indicate regressions.

This test will compare the current output of abiword --to=abw <file> against a stored version of the output. The comparison is done using the 'diff' command.

Note that whenever the raw diff test reports a change, it does not necessarily mean that an error has been found. Several different cases can be identified:

• an importer has been changed, for example to fix a bug or to add a new feature. A change reported by the raw diff test should be examined closely to determine if the change includes ONLY the expected result of the change. If it includes some other unexpected/unwanted changes in the output as well, a regression has been found.

• AbiWord outputs awml differently, for example because of an internal change, or a change in the awml exporter. The raw diff test will now report changes for possible all exported files. In this canse no real harm has been done, and thus the change should not be considered as an error.

Whenever the awml output changes as expected, the raw output for every file in the test set should be regenerated and stored. This process is automated, and can be done simply by calling:

	./regenerate_raw.pl

After the raw output has been updated, the changes to the .raw. files should be stored in the repository.

NOTE: this process will take a while, as a totally new clean build for each branch is created (which is slightly dumb, but the easiest way to make sure the raw files are created with the correct abiword version.

This test checks for memory errors during the conversions, as well as for memoryleaks.

In the various subdirectories (abw, doc, rtf, etc.) of ./impexp, files can be added at will. Make sure to also add them to the regression.in file located in each directory (note: you can disable specific files from being tested by commenting them out with a '#').

When you finished adding new files, enter the impexp directory and run:

	./regenerate_raw.pl

Commit the results to cvs if you want the new files to be part of the 'official' testruns.

NOTE: make sure to have executed ./bootstrap.pl at least once before running ./regenerate_raw.pl. When you have executed the regression test before, bootstrap.pl has already been executed for you.

<not implemented at the moment>

This test compiles both abiword and abiword plugins, and checks if there are no compiler warnings. The presence of warnings could point out a true problem, so it is important to keep the code compiling warning free (which can be hard, given the different compilers out there).