[Scons-dev] Progressing D tool support

[Gary Oberbrunner]

On Thu, Aug 9, 2012 at 2:23 PM, William Deegan
<bill at baddogconsulting.com> wrote:

>> One of the reasons for not merging the pull request was a lack of new

>> tests, which is fine in a sense and impossible to comply with in

>> another. The changes were all about versions of D installed. I am not

>> entirely sure there are tests that can be written to specifically vet

>> the changes made. Moreover I suspect no-one other than me has D

>> installed – I have a deep suspicion that the D tests are not run by the

>> CI.


Having D (some version) and testing against it is of course the ideal,
and we should have at least one machine run some live D tests.  But,
using MSVC as an example, a lot of the MSVC tests don't actually run
the compiler.  They just unit-test various parts of the
infrastructure: do the various $CCFLAGS and so on actually get through
to the command line?  Does the Tool actually run, and populate the env
correctly?  Can it find what it needs in the registry?  (MSVC does
that by having a fake known-good registry setup, and mocking things up
so SCons uses the fake registry, but that's a pretty complex test
rig.)  I'm just saying there's probably plenty of simple testing that
can be done without any actual D installed, if you're willing to do
some mocking.

-- 
Gary