All,

On Feb 4, 2013, at 11:31 AM, "Kenny, Jason L" <[email protected]> wrote:

> 
> I am not going to say this is an easy problem to solve ( dealing with this in 
> Parts, all the time)
> 
> Depending on how you view the "tool" to work I find that I need to test did 
> tool X:
> 1) correctly detect that it does not exist
> 2) detect that it does exist
> 3) given that it does exist, is the environment correctly setup to build with 
> the given tool.
> 4) I am a little more complex, as I have Parts support different 
> version/targets platform combination. So testing different version selection 
> is needed for me as well. ( My current headache is how to test for certain 
> SDKs such as the NDK to make sure the test run only if this exists, without 
> having to set up some super path before the tests runs, or binding to a 
> certain version of the NDK) 
> 
> If you setup for the different D tools is like the C++ cases, ie we have 
> intelc, msvc, gcc, etc... then each case can be tested independently from 
> each other. It could be done on a single machine with the D tools installed. 
> If this is a single D tool that deal with all cases. That would be hard to 
> test, it would probably need lots of VM, so you could control which D 
> compiler is being used.

Adding to the complexity is that MS licenses are not free..

> 
> I am only thinking that it might not be so hard given we have different tool 
> files for the different D compiler cases. One machine for a given platform ( 
> windows, linux, mac, etc) can test the exists cases, while on can deal with 
> all cases of tools not existing.

Would it make sense to have the test infrastructure take an environment 
variable on tools to not initialize?
And/or for SCons itself, perhaps also a command line argument?
This would simplify the # of VM's needed to test the possible or reasonable 
combinations?

-Bill
_______________________________________________
Scons-dev mailing list
[email protected]
http://two.pairlist.net/mailman/listinfo/scons-dev

Reply via email to