[catalogue] Fix VMSA register size and type mismatch

[diaolo01]

This PR adresses register type mismatches in aarch64-vmsa tests:

• fix pteval_t vs int register type mismatch
• fix register size mismatches

[relokin]

Thanks @diaolo01.

@maranget independently of the specifics of this change, do you have long term logs for aarch64-VMSA? Should we aim to preserve hashes here?

[relokin]

Along with this we should remove LDR+32. But I am not sure these set of tests should be here in the first place. I think these should be in herd/tests/instructions - if there is no such test there already.

[maranget]

The catalogue is also the test base for the web interface. We may want simple tests such as those to appear un the web interface.

[diaolo01]

I opted for keeping this test given Luc's comment and made the registers 64 bit instead to differentiate them from 32 bits tests.

[relokin]

FWIW, the feedback I get from users is that web and in particular the aarch64-VMSA is overwhelming and not as well curated as it could be. I think having 100s of tests doesn't help and having both a 64bit LDR and a 32bit LDR only makes the situation worse. If we want to maintain these tests for testing, and I am not sure we need to, I think we should create a shelf.py for regressions and a separate shelf.py for the web.

[maranget]

Fair enough. If I am not mistaken the regression test driver can take an index file as argument.

[maranget]

Thanks @diaolo01.

@maranget independently of the specifics of this change, do you have long term logs for aarch64-VMSA? Should we aim to preserve hashes here?

Hi @relokin abd @diaolo01. There are no long term logs that depend on any test from the github. All runs are performed on copies taken a long time ago. So it is safe to change logs here as long as references are also changed, when appropriate.

[maranget]

Hi again @diaolo01 and @relokin. I have started the process of editing the tests, with the following aims

• Suppress herd7 errors (there are some since PR [herd] Strict mem sizes #1588).
• Suppress litmus7 warnings.
• Remove all Hash metadata, as they are useless.

Then I noticed that some tests have already been fixed by merged PR #1762. Would you please rebase?

[maranget]

I have already fixed some tests here, commit 35b45c4. May I suggest cherry-picking this commit?

[diaolo01]

Thank you @maranget, looks like we did some of the same work in parallel. Should I cherry-pick the rest of the changes on top of this commit?

[maranget]

If you mean that my commit has more changes than yours, yes please proceed.

[maranget]

Thank you @maranget, looks like we did some of the same work in parallel.

Hi @diaolo01. To avoid this I won't work on this PR unless you tell me you have finished...