[Scons-dev] Tests

William Deegan bill at baddogconsulting.com
Thu Sep 6 12:10:50 EDT 2012


Russel,
On Sep 5, 2012, at 8:11 PM, Russel Winder <russel at winder.org.uk> wrote:


> On Tue, 2012-09-04 at 17:29 +0200, Dirk Bächle wrote:

> [….]

>> in my intermediate "test framework" project on Bitbucket I supported the

>> special "sconstest-*.py" syntax for marking tests.

>> When integrating the support for testing external Tools into the SCons

>> core, I skipped this naming rule for the benefit of staying fully

>> compatible with the regression tests so far, and not having to introduce

>> another filename syntax.

>

> This naming policy is actually extremely helpful, it is a pity is does

> not work in the core. I have many modules that are support modules for

> tests, they are not tests themselves but are just shared code to avoid

> replication. Without a naming scheme you cannot do any modules that

> aren't actually test modules.

>

> Or perhaps there is a place for modules that are on the PYTHONPATH, not

> actual tests but are modules accessible to tests.

>

>> At the moment, meaning latest revision in "default", the test framework

>> supports the skipping of whole

>> folders (like, for a supporting Python package) by adding a

>> "sconstest.skip" file. A list of single files to skip can also be

>> specified in a file named ".exclude_tests (see

>> QMTest/test-framework.rst, sect. "Finding Tests").

>>

>> Does this help already, or do you need a different functionality?

>

> This is getting close. The issue is whether the directory is on the

> PYTHONPATH or can trivially be put on the PYTHONPATH without egregious

> hacking.

>

> For now I will create a Support directory for the test/D hierarchy and

> edit sys.path to add this to the path – but this is exactly egregious

> hacking :-(


Is that really necessary.
We have the MSCommon dir in which there are no tests, but just support modules for the various tests for MSVC,MSVS,..

They are not run as tests.
-Bill


More information about the Scons-dev mailing list