Enable tests failing due to missing array ctor and lbound index checks.

[lateralusX]

Fix mono/mono#19494 adds missing array ctor needed by several tests currently failing as described by #34068.

This PR fixes an additional problem hit by the IL array tests using an lbound != 0. Current Mono codegen for Set/Get/Address array methods didn't adjust index with regards to lbound causing failures in several of the array tests. NOTE, this scenario should only be hit when directly calling Set/Get/Address using IL on an array with an lbound != 0. It shouldn't affect regular szarray's at all since that code paths should never be used and the codegen currently looks up the lbounds unconditionally in the array, so if that code path ever gets hit by a szarray, it will imdediate crash, so shouldn't be causing any performance impact on szarrays. There is also an additional check on the array type as well as an additional precaution, so only arrays of type MONO_TYPE_ARRAY, hitting codegen for Set/Get/Address array method calls will adjust index based on lbound.

This PR fix the lbound index codegen for that specific code path and re-eanbles the tests currently listed in #34068.

[ghost]

Tagging subscribers to this area: @ViktorHofer
Notify danmosemsft if you want to be subscribed.

[danmoseley]

Poor bot, I guess it saw ".targets"

[lateralusX]

Depending on #34054 to be merged before CoreCLR test suites will run using Mono on CI.

[lateralusX]

@naricc CoreCLR tests are now running on Mono runtime, correct? If so I guess it's time to rebase this PR and validate the results when reenable the array ctor tests on CI.

[naricc]

@lateralusX Yes, the runtime tests (I'm trying to stop calling them the CoreCLR tests, since they now are testing multiple runtimes and it gets confusing :) ) are now in CI. You will see them as "Mono Pri0 Runtime Tests Run OSX x64 release". I am working on standing up CI lanes for other architectures/modes (like interpreter).

[lambdageek]

@naricc could you resolve the conflicts in this PR

[naricc]

@lambdageek Ok; on my list.

[stephentoub]

There are conflicts. Mind rebasing? Thanks.

[lateralusX]

Investigate this a little further, we now get passed the error fixed by mono/mono#19494, but several of these tests also use constructions to test lbound != 0 and then access the array using indexes inline with the new low and upper bounds. On Mono this will fail with 1 dim arrays since we don't adjust index based on lbound != 0. This needs to be fixed before enable the tests or we will still get failures.

[stephentoub]

Thanks. Is there an open issue tracking that?

[stephentoub]

(And should this be closed in the interim?)

[lateralusX]

I believe the only one we currently have is the one referenced in this PR, #34068, but that highlights another problem that is solved (missing constructor). I will look into a potential fix for the other things not working in the tests, but depending on the outcome of that fix and perf impact, I could add that fix into this PR and re-enable tests as part of complete fix. If we come to the conclusion that the fix adds to much perf overhead, I believe we also decided that we won't fix the failing tests, so then we can go ahead and close both this PR and the original issue.

[lateralusX]

I will rebase this PR, add additional fix into this PR and re-enable all the tests tomorrow morning.

[lateralusX]

@naricc I take over this PR again since it needs some additional fixes before all tests pass (not enough to just enable them).

[lateralusX]

Tests pass on Windows x64, so expected at least OSX x64 to get same results, but test runs looks like the additional codegen is not hit, since failures looks close to what happens when the lbound is not adjusted. Need to investigate the tests on none Windows platforms.

[lateralusX]

Failures was not in JIT, but interpreter config, updated with similar adjustments done to JIT in interpreter.

[lateralusX]

There is also a missing OP code on arm64 causing issue to convert float -> int, adding simliar conversion for OP_FCONV_TO_I as done on amd64, arm and x86.

[lateralusX]

Down to 3 failures, only interpreter lane on _il_relgcarr, _il_relint32_m1, _il_dbgint32_m1, so these three cases must hit some other code path in interpreter compared to rest, will investigate.

[lateralusX]

Down to one failing tests in mono/mono, apparently we take one internal shortcut when creating a MONO_TYPE_ARRAY with lower bound of 0 and don't create the runtime lbound struct, so we need to add condition in codegen for that case.