CI is red again

[hmelder]

[fredkiefer]

Hi, this is great thank you for this clever fix. But this also raises a new question. GNUstep code as well as application code is full of IMP caching and my impression now is that that code also would be affected of the sam issue. Should we rewrite all that code to use the proper way to get the method implementation and advise users to do the same?

[hmelder]

Hi, this is great thank you for this clever fix. But this also raises a new question. GNUstep code as well as application code is full of IMP caching and my impression now is that that code also would be affected of the sam issue. Should we rewrite all that code to use the proper way to get the method implementation and advise users to do the same?

I think this is the wrong issue ^^

[rfm]

Yes, it probably is a bit of a misunderstanding, but I think the point is related to this issue:

I'm away from home at the moment, so this is from memory, but when cacheing IMPs we want to be sure we are looking them up correctly, so it is probably worth checking that the NSObject -methodForSelector: implementation is doing the right thing.

…

[triplef]

These are the two remaining failures on Windows MSVC (everything else is green):

base/NSURLSession/simpleTaskTests.m:
Failed file:  simpleTaskTests.m aborted without running all tests!

base/NSURLSession/uploadTaskTests.m:
Failed file:  uploadTaskTests.m aborted without running all tests!

They both fail with this error:

Uncaught exception NSInternalInconsistencyException, reason: failed to load HTTPServer.bundle

It looks like the HTTPServer helper is not actually being built: running the tests just creates the empty folders HTTPServer.bundle and obj/HTTPServer.obj. Is bundle.make maybe not supported on Windows?

[hmelder]

I'll take a look at it tomorrow :)