[cling] Desugar UsingType in GetFullyQualifiedType

[devajithvs]

The change in getFullyQualifiedType() in the commit llvm/llvm-project@af27466 added explicit desugaring of UsingType:

  if (isa<UsingType>(QT.getTypePtr())) {
    return getFullyQualifiedType(QT.getSingleStepDesugaredType(Ctx), Ctx,
                                 WithGlobalNsPrefix);
  }

This has been there since LLVM14 and we could disentagle this from the upgrade to LLVM22.

This will cause SG::sgkey_t to be resolved to its underlying type uint32_t when GetFullTypeName() is called, rather than preserving the typedef alias name as before.

This test was added in: #19615, should fix: #12698

This will be the default behaviour (in LLVM22) after #21658. So decoupling this change from the upgrade here.

This Pull request:

Changes or fixes:

Checklist:

• tested changes locally
• updated the docs (if necessary)

This PR fixes #

[github-actions]

Test Results

    23 files      23 suites   3d 15h 3m 5s ⏱️
 3 874 tests  3 851 ✅ 0 💤 23 ❌
78 807 runs  78 784 ✅ 0 💤 23 ❌

For more details on these failures, see this check.

Results for commit 0abc723.

♻️ This comment has been updated with latest results.

[vgvassilev]

Is this is what we really want? I am not sure if we should be losing that type sugar but @pcanal should be able to tell...

[devajithvs]

This is what upstream does. We maintain a separate GetFullyQualifiedType implementation anyway. I was trying to reduce the divergence with the upgrade.

[vgvassilev]

This is what upstream does. We maintain a separate GetFullyQualifiedType implementation anyway. I was trying to reduce the divergence with the upgrade.

Yes, I agree with that, let's see what @pcanal has to say if that's not breaking something in i/o conceptually.

[pcanal]

Thanks.

[hahnjo]

To mirror what I wrote on Mattermost, after looking at the test change, I'm not so sure anymore: SG::sgkey_t is the full type name of data member m_sgkey. It aliases to uint32_t, which then aliases to unsigned int

[devajithvs]

Yes, we should probably get the canonical type instead of just going one layer down.

[vgvassilev]

Yes, but this is using then the wrong facility..

[devajithvs]

I see, thanks for the pointer. I understand now that GetFullTypeName() should preserve the alias spelling as it was written.

I will close this PR, keep the existing m_sgkey test as-is, and add a separate test case for UsingType (same behavior as TypedefType) in LLVM22 PR.

The problem here was that, even though top-level node here is a TypedefType:

TypedefType 0x5bca02c32f60 'SG::sgkey_t' sugar
  ...
  UsingType 0x5bca02c32490 'std::uint32_t' sugar
  ...
    UsingShadow 0x5bca02c32280 'uint32_t'

getAs<UsingType>() traverses through the type sugar until it finds the underlying UsingType node, here returns the UsingType declaration for std::uint32_t.