Testing

[Granville Barnett]

Hi everyone,

How do people isolate and test their custom additions to the SSCLI? From what I have seen so far performing *clean* testing on components that interact heavily with those already within the SSCLI is either infeasible or just down right messy.

By testing I mean in the form of unit tests.

Anyone have any suggestions or solutions which worked best for them?

Cheers,

Granville

[XWang]

What is the *clean* test? I am quite interested...

When I make changes to sscli, I isolate the files in another place out
of the sscli, and override the sscli when i make changes to sscli,
Also another place to make my own tests.
I think you must have known that the unit testing are all in the tests
folder

On 1月7日, 上午9时22分, "Granville Barnett"

[Granville Barnett]

Yep, I knew about those tests.

But the vast majority of those tests verify at an abstract level that it *works*. I am on about more atomic tests that can verify smaller pieces of functionality, e.g. if I modify the GC algorithm and add some other heuristic to it and I want to test that this heuristic is satisfied for a particular situation and its state is correct I would rather do that in a unit test style approach rather than using the approach in the SSCLI. 

I'm not adding anything to the GC I'm just using it as an example.

Also I have found some of the tests in the SSCLI to be not that good as I have had builds and their associated tests pass yet I know that my modifications to the C# compiler would yield breakages. 

That last point is probably the source of my intent to adopt an approach which I know works.

Granville

[XWang]

Sounds cool, However, Adding a algorithm a good approach. What kind of
Modifications have you made, would you like to share it.
To be honest, i just add a cil opcode yet. but quite interested to
take further step.

On 1月7日, 上午11时31分, "Granville Barnett"

> > > Granville- 隐藏被引用文字 -
>
> - 显示引用的文字 -

[Granville Barnett]

I've been looking at modifying various components within the SSCLI from the JIT to the C# compiler. I'm just trying to figure out the best testing approach as it seems to have been neglected in other work I've seen built upon the SSCLI. 

When I have something to share I will do ;-)