3.9 KiB

Raw Blame History

pdf2htmlEX tests

This directory contains a collection of python3 unittests of the output of pdf2htmlEX.

The graphical output of pdf2htmlEX can be tested both locally and remotely using Selenium and the Pillow Python Imaging Library.

The browser tests use Selenium to take a screenshot of a FireFox browser's rendering of the pdf2htmlEX output for a given pdf file and compares that image to an image of the previously saved reference html.

Tests which are currently failing:

browser_tests/text_visibility At the moment clipping has been broken and needs to be fixed. Rerun runLocalBrowserTests and use the compareTestImages for the test_visibility test to see the problem.

Running tests

There are three bash scripts which automate the running of a given collection of tests:

runLocalTests runs a simple collection of tests which do not require Selenium or a browser.

  ./runLocalTests

runLocalBrowserTests runs a more complex collection of tests which require Selenium, a FireFox browser, as well as a 'virtual frame buffer' (Xvfb) to be installed.

  ./runLocalBrowserTests

runRemoteBrowserTests runs the same complex collection of tests as run by runLocalBrowserTests but this time using 'Sauce Connect'. (At the moment this is not fully implemented or (re)tested)

  ./runRemoteBrowserTests

In order to run these tests, you must have the correct testing software installed locally. To do this you can run the command:

  ./installAutomaticTestSoftware

Understanding browser test failures

If any of the automatic browser tests fail then you might want to manually view the PNG images for a given test using the command:

  ./compareTestImages <<testNam>>

This command opens the three PNG images associated with a given failed test so that you can manually compare the new output (*.out.png), the reference output (*.ref.png) and an image of the 'difference' between the two images (*.diff.png). To pass, the 'difference' image must be completely black.

Usually it will be obvious that the newer version of pdf2htmlEX has only slightly moved various image elements. Any such tests can be made to pass by updating the reference html using the tool:

  ./regenerateTestHtml <<testName>>

This command will regenerate the reference html for the specifed test.

All of these manual comparison tools require additional software which can be installed using the command:

  ./installManualTestSoftware

OLD README contents:

Dependencies

python2 and packages
- Python Imaging Library
- Selenium
- unittest
Firefox
- firefoxdriver

Usage

Run all tests:
- python test_output.py
- python test_local_browser.py
Environment variables:
- export P2H_TEST_GEN=1 to generate new reference files (when done, unset P2H_TEST_GEN)
- export P2H_TEST_REMOTE=1 to test different browsers using Sauce Labs
  - Install sauceclient for Python
  - Set correct values for SAUCE_USERNAME and SAUCE_ACCESS_KEY
  - Setup a HTTP server at / on port 8000
  - Enable Sauce Connect
  - See .travis.yml as an example
  - python test_remote_browser.py

Add new test cases

Make sure you have the proper copyrights.
Using meaningful file names, a description of the file, or issueXXX.pdf.
Make each test case minimal:
- One page only, unless the test case is about multiple pages.
- Grayscale only, unless the test case is about colors.
- Remove unnecessary elements.
[Optional] Include the source files that the PDF file is generated from.
Add the new PDF file to the correct folder in test/, and add a new function in the corresponding Python file
Run P2H_TEST_GEN=1 test/test.py test_issueXXX to generate the reference, assuming that the new function is called test_issueXXX

3.9 KiB Raw Blame History