Create new test system for name resolution 2.0

[powerboat9]

This should allow us to more easily test and record progress on name resolution 2.0

[powerboat9]

Note that, until name resolution 2.0 improves sufficiently, a decent fraction of new tests after this is merged (?) will have to be entered into the exclude file.

[CohenArthur]

@powerboat9 I like the implementation and I think this feature makes sense, but I'm not sure if it's going to make our life easier in the long run. I'm in favor of merging it if the rest of the community thinks it is a good idea. it does add a burden which is maintaining and revisiting that exclude file. but the alternative is also annoying, as we'd need to add a copy of existing testcases with -frust-name-resolution-2.0 added as an argument...

maybe we can do something like keep the nr2 folder, run it with different dejagnu options but only keep symlinks to testcases that pass under NR2.0?

[P-E-P]

maybe we can do something like keep the nr2 folder, run it with different dejagnu options but only keep symlinks to testcases that pass under NR2.0?

AFAIK symlink wouldn't work when running tests on windows.

On one hand I really like this PR's idea but maintaining this list of file looks horrible. I'm in favor of merging this for now, we could maintain that list until the nr2 is stable and working as expected, we could then come back to this decision.

[powerboat9]

Thoughts on merging this? Should I make some tweaks?

[P-E-P]

Thoughts on merging this? Should I make some tweaks?

To be honest, multiple peoples are not too keen on the implementation (and so am I). The idea is neat in theory but name resolution 2.0 should be merged shortly once we have fixed the final missing test and this PR brings a burden as we have to update the list. So there are two points to figure out:

• Do we really want to maintain a list of name resolution 2.0 test like this ? Isn't there a way to automagically add the tests to their own batch
• Is it too late to separate nr 2.0 tests from the others ? I have one final test to fix before creating a new PR for nr 2.0, this does not mean we will turn it by default as soon as it is merged but the transition should happen fairly early.

Recently I've introduced some tests subdirectories (macros), but in your PR you chose a different approach, with a text list, is there any reason you chose not to got with a subdirectory ? I surely miss a detail somewhere.

[powerboat9]

Ah, I didn't realize nr2.0 was so close to completion. The idea was that every compile test for non-nr2.0 would automagically run with nr2.0 as well, except for those in the exclude list. As nr2.0 got closer to completion we could slowly remove tests from the exclude list, and once it was empty we'd be able to enable nr2.0 for all tests and revert this PR. If nr2.0 is about to be completed this would be moot.

[powerboat9]

Where is the development on nr2.0 taking place? I think I might be a bit out of the loop

[P-E-P]

Where is the development on nr2.0 taking place? I think I might be a bit out of the loop

I need to open a PR but I've been working on top of #2940. If you want to get an exact idea on what is working/missing you could compile my branch https://github.com/P-E-P/gccrs/tree/allow-use-before-items-being-used-continued. Right now one test is still failing (use duplication, name_resolution21.rs).

[powerboat9]

When I try to force-enable nr2.0 on allow-use-before-items-being-used-continued (by replacing flag_name_resolution_2_0 references with 1) I get a bunch of test errors. Are you sure there's only one non-working test?

[P-E-P]

When I try to force-enable nr2.0 on allow-use-before-items-being-used-continued (by replacing flag_name_resolution_2_0 references with 1) I get a bunch of test errors. Are you sure there's only one non-working test?

You're right, I didn't apply the correct patch. Looks like most of those fails are due to an ICE (but not all of them):

• rust/resolve/rust-late-name-resolver-2.0.cc:206
• rust/resolve/rust-late-name-resolver-2.0.cc:157
• rust/hir/rust-ast-lower-base.cc:655
• rust/resolve/rust-name-resolver.cc:661
• We also have a segfault somewhere

The situation is worse than I initially thought.

[P-E-P]

I can't fix those in the upcoming PR as it is becoming too big. And there is a lot of work remaining, we should merge this PR.

[P-E-P]

Wait I thought the error was due to some github queue shenanigans but it doesn't. @powerboat9 would you mind rebasing this PR on master ?

[powerboat9]

Yep, looks like it needs some tweaks after some tests were moved around. Will need to be adjusted to handle sub directories properly.

[powerboat9]

Alright, this should do it. We should also now get XPASS reports when tests in the exclude list no longer need to be excluded.