Lack of specifications for test cases

Bug #1733918 reported by Yann Pouillon
6
This bug affects 1 person
Affects Status Importance Assigned to Milestone
Siesta
New
Undecided
Unassigned

Bug Description

I'm currently carrying out a testing campaign on various builds of the trunk of SIESTA (revno 646). I want to automate everything as much as possible, in order to avoid human mistakes and schedule the tests when the computer is most available.

In most cases, I can have a generic shell-script calling 'make' for individual tests. However, I'm not able to decide what to do for some test cases or have to manually change the script to run it successfully. I will use a few examples here to make questions about the design of the test suite.

Example 1: TDDFT_Ge contains 2 input files, TTDFT_Ge1.fdf and TDDFT_Ge2.fdf. Which one should be run? Does the order matter?

Example 2: The input files of the flos_h2o_* tests are called h2o_*.fdf, which means I cannot automatically check their presence by using the directory name of the test. Is this voluntary?

Example 3: The TRANSIESTA tests have the same issue as example 2, in addition to be in subdirectories of TranSiesta-TBTrans/. I can deal with directory hierarchies quite easily, but their combination with changes in input file names makes it a bit cumbersome.

I would like to discuss design principles for test cases, in order to determine if we can converge on a minimum set of simple criteria when writing new tests. This will be particularly useful to me when adding the tests for the hybrids.

Tags: testsuite
Revision history for this message
Nick Papior (nickpapior) wrote :

Example 1.
When running make in the TDDFT_Ge sub-directory it will run _1 first, then _2. This should be controlled by the test_tddft.mk file. However, I agree it could be updated for clarity.

Example 2.
I will go through the tests in 4.0 and update names to be consistent.

Example 3.
Currently this is limited because one has to compile transiesta as a separate executable. Hence those tests are sort of different. Probably this would change if I get the time to make transiesta a part of siesta (so one does not need to different executables).
In that case then we could prefix transiesta tests with TS_, just like TDDFT.

Revision history for this message
Yann Pouillon (pouillon) wrote :

Thanks. Clarity is definitely what is needed. A few simple recommendations on how to design test cases would likely be enough.

Revision history for this message
Nick Papior (nickpapior) wrote :

Example 2 has just been fixed in 4.1, this will propagate to trunk (will let you know). Once merged you can pull into your branch.

Revision history for this message
Nick Papior (nickpapior) wrote :

Feel free to merge with trunk now.

To post a comment you must log in.
This report contains Public information  
Everyone can see this information.

Other bug subscribers

Related blueprints

Remote bug watches

Bug watches keep track of this bug in other bug trackers.