RFR: 7903600: Add more tests for unsupported types

[Maurizio Cimadamore]

On Mon, 4 Dec 2023 20:44:14 GMT, Jorn Vernee <jvernee at openjdk.org> wrote:

>> Add more tests for unsupported types. See the new TestUnsupportedTypes.java
>> 
>> While working on this, I also noticed several cases of dead code: `FunctionalInterfaceScanner` is completely unused, and `InMemoryJavaCompiler::jfoFromByteArray` and `InMemoryJavaCompiler::jfoFromString` (and its caller in `OutputFactory`) are unused as well. I've removed them in this PR.
>> 
>> Changes in this PR:
>> - I'm passing in the error stream PrintWriter that we create in the test all the way down to UnsupportedFilter, so that it can write to it instead of `System.err`. Then we can check what was written in the test. 
>> - I've added warning print outs for the different reasons why we are skipping declarations. Some of these were missing.
>> - I've removed a check for ValueLayouts that are larger than 8 bytes in `visitVariable`. If the type of a variable is an unsupported primitive, we already filter it out earlier on in the method, so there's no need to check again.
>> - I've added a missing `Skip.with` in `visitVariable`, so that we will filter out (global) variables with unsupported function types.
>> - I've added skipping of typedefs when they have an unsupported type or a type that does not have a layout. For the latter, I've also had to remove a test case that checked if we generated a typedef class for an undefined struct type. Since the struct is undefined, I believe we should not generate a typedef binding class for that case.
>> 
>> The new test covers all the unsupported type cases that are handled by TestUnsupportedTypes.
>
> test/testng/org/openjdk/jextract/test/toolprovider/Test8245767.java line 50:
> 
>> 48:             Class<?> fooCls = loader.loadClass("Foo");
>> 49:             assertNotNull(fooCls);
>> 50: 
> 
> This is the problematic test case that I believe to be testing for behavior that we do not want.

I've seen that yesterday, and was equally puzzled by the "expected behavior"

-------------

PR Review Comment: https://git.openjdk.org/jextract/pull/152#discussion_r1415308046