[Verifier] Allow DW_TAG_class_type/DW_TAG_union_type to have no filename

[MaskRay]

Summary

clang/lib/CodeGen/CGOpenMPRuntime.cpp synthesized union
(distinct !DICompositeType(tag: DW_TAG_union_type, name: "kmp_cmplrdata_t", size: 64, elements: <0x62b690>))
does not have meaningful filename/line number.

D94735 dropped the previously arbitrary and untested filename/line from the
union and caused a verifier error here.

This fixes check-libarcher failures.

Intended for main & release/12.x

[protze.joachim]

I just compared the debug information generated for a build from 7630520ae + this patch with a build from 7630520ae + revert 189f3111. (7630520ae was just the top of main, when I did the bisecting, so I already had a complete build)

I diffed the generated .debug_info section for all binaries generated with check-libarcher. I used readelf -wi to extract the .debug_info section. The only difference is really just the DW_AT_producer attribute which is clear because the clang info contains the git hash.
So, at least for this small collection of OpenMP programs we don't loose debug information by 189f3111 + this patch.

This is all the information available for kmp_cmplrdata_t in the old and the new build:

<1><6c2e3>: Abbrev Number: 29 (DW_TAG_union_type)
   <6c2e4>   DW_AT_name        : (indirect string, offset: 0xddd4): kmp_cmplrdata_t
   <6c2e8>   DW_AT_byte_size   : 8

[aprantl]

I suppose there is no good reason to require a file name. In other places we work around this by creating a file with the name "<compiler-generated>" but this doesn't add a lot of information. Do the various backends in AsmPrinter (DWARF+CodeView) accept null Files on these nodes?

[aprantl]

In any case, I'm fine with this change, or with changing clang to create a special artificial file to stick the compiler-generated declarations into.

[dblaikie]

In D96212#2549513, @aprantl wrote:

I suppose there is no good reason to require a file name. In other places we work around this by creating a file with the name "<compiler-generated>" but this doesn't add a lot of information. Do the various backends in AsmPrinter (DWARF+CodeView) accept null Files on these nodes?

hmm - got any pointers to places where we create a file named "<compiler-generated>"? Wouldn't mind taking a look at those/maybe changing them to not have a name.

[dblaikie]

In D96212#2549526, @aprantl wrote:

In any case, I'm fine with this change, or with changing clang to create a special artificial file to stick the compiler-generated declarations into.

(oh, and I'm fine with it if/since Adrian is)

llvm/lib/IR/Verifier.cpp

if (N.isVector()) {
     AssertDI(Elements.size() == 1 &&
              Elements[0]->getTag() == dwarf::DW_TAG_subrange_type,
              "invalid vector, expected one element of type subrange", &N);
   }
 
   if (auto *Params = N.getRawTemplateParams())
     visitTemplateParams(N, *Params);
 
-  if (N.getTag() == dwarf::DW_TAG_class_type ||
-      N.getTag() == dwarf::DW_TAG_union_type) {
-    AssertDI(N.getFile() && !N.getFile()->getFilename().empty(),
-             "class/union requires a filename", &N, N.getFile());
-  }
-
   if (auto *D = N.getRawDiscriminator()) {
     AssertDI(isa<DIDerivedType>(D) && N.getTag() == dwarf::DW_TAG_variant_part,
              "discriminator can only appear on variant part");
   }
 
   if (N.getRawDataLocation()) {
     AssertDI(N.getTag() == dwarf::DW_TAG_array_type,
              "dataLocation can only appear in array type");