[DebugInfo] Fix reading location tables headers of v5 units in DWP.

[ikudrin]

Summary

This fixes the reading of location lists headers for compilation units in package files by adjusting the reading offset according to the corresponding record in the unit index. This is required for DW_FORM_loclistx to work.

[jhenderson]

Looks good to me, although I'm not fully up to speed on .debug_loclists to be 100% certain, so might be worth a second opinion.

What did the code do before? Parse the first one repeatedly?

llvm/lib/DebugInfo/DWARF/DWARFUnit.cpp

541–542

Maybe these shouldn't be auto? Certainly, C looks like it should be a number (e.g. uint64_t), but it obviously isn't.

[ikudrin]

In D77146#1954186, @jhenderson wrote:

Looks good to me, although I'm not fully up to speed on .debug_loclists to be 100% certain, so might be worth a second opinion.

Thanks!

What did the code do before? Parse the first one repeatedly?

The code is OK for a DWO case, but is not ready for DWP. The support for DWARFv5 DWP files is added in D75929, so this code also needs to be updated.

llvm/lib/DebugInfo/DWARF/DWARFUnit.cpp

541–542

getOffset() returns const DWARFUnitIndex::Entry::SectionContribution *. Not that obvious, I agree. Do you think it would be better to rename C to, say, Contrib or Contribution?

[jhenderson]

llvm/lib/DebugInfo/DWARF/DWARFUnit.cpp

541–542

I think the issue actually is the function name: getOffset probably wants renaming to something like getContribution. Contrib is probably fine for the variable name (though I'm not sure if using`auto` here isn't a violation of LLVM's coding standards, regardless of the variable or function name).

[dblaikie]

llvm/lib/DebugInfo/DWARF/DWARFUnit.cpp

541–542

Yeah, I don't mind the rename - either before or after this change. I'm pretty OK with "auto" given the very short scope & pointer-ness (making it obvious what boolean testing the thing does, etc)

[ikudrin]

• C -> Contrib.

I've created D77302 for renaming getOffset functions.

[ikudrin]

• Rebase to the tip.
• const auto *Contrib -> auto *Contrib.

[dblaikie]

Seems good - thanks!

llvm/lib/DebugInfo/DWARF/DWARFUnit.cpp

542

I'd leave the const here

llvm/lib/DebugInfo/DWARF/DWARFUnit.cpp

if (LocSection->Data.size()) {
       DWARFDataExtractor Data(Context.getDWARFObj(), *LocSection,
                               isLittleEndian, getAddressByteSize());
       if (Offset < HeaderSize)
         return createStringError(errc::invalid_argument,
                                  "did not detect a valid"
                                  " list table with base = 0x%" PRIx64 "\n",
                                  Offset);
       Offset -= HeaderSize;
+      if (auto *IndexEntry = Header.getIndexEntry())
+        if (const auto *Contrib = IndexEntry->getOffset(DW_SECT_LOCLISTS))
+          Offset += Contrib->Offset;
       if (Error E = LoclistTableHeader->extract(Data, &Offset))
         return createStringError(errc::invalid_argument,
                                  "parsing a loclist table: " +
                                      toString(std::move(E)));
     }
   }
 
   // Don't fall back to DW_AT_GNU_ranges_base: it should be ignored for

llvm/test/DebugInfo/X86/dwp-v5-loclists.s

 ## The test checks that v5 compile units in package files read their
 ## location tables from .debug_loclists.dwo sections.
 ## See dwp-v2-loc.s for pre-v5 units.
 
 # RUN: llvm-mc -triple x86_64-unknown-linux %s -filetype=obj -o - | \
 # RUN:   llvm-dwarfdump -debug-info -debug-loclists - | \
 # RUN:   FileCheck %s
 
 # CHECK:      .debug_info.dwo contents:
 # CHECK:      DW_TAG_compile_unit
 # CHECK:      DW_TAG_variable
 # CHECK-NEXT:   DW_AT_name ("a")
 # CHECK-NEXT:   DW_AT_location (0x{{[0-9a-f]+}}:
 # CHECK-NEXT:     DW_LLE_startx_length (0x0000000000000001, 0x0000000000000010): DW_OP_reg5 RDI)
+# CHECK:      DW_TAG_variable
+# CHECK-NEXT:   DW_AT_name ("b")
+# CHECK-NEXT:   DW_AT_location (indexed (0x1) loclist = 0x{{[0-9a-f]+}}:
+# CHECK-NEXT:     DW_LLE_startx_length (0x0000000000000005, 0x0000000000000020): DW_OP_regx RDI)
 
 # CHECK:      .debug_loclists.dwo contents:
 # CHECK:      locations list header:
 # CHECK:      locations list header:
+# CHECK:      offsets:
 # CHECK:      0x{{[0-9a-f]+}}:
 # CHECK-NEXT:   DW_LLE_startx_length (0x0000000000000001, 0x0000000000000010): DW_OP_reg5 RDI
+# CHECK:      0x{{[0-9a-f]+}}:
+# CHECK-NEXT:   DW_LLE_startx_length (0x0000000000000005, 0x0000000000000020): DW_OP_regx RDI
 
 .section .debug_abbrev.dwo, "e", @progbits
 .LAbbrevBegin:
     .uleb128 1                          # Abbreviation Code
     .uleb128 17                         # DW_TAG_compile_unit
     .byte 1                             # DW_CHILDREN_yes
     .byte 0                             # EOM(1)
     .byte 0                             # EOM(2)
     .uleb128 2                          # Abbreviation Code
     .uleb128 52                         # DW_TAG_variable
     .byte 0                             # DW_CHILDREN_no
     .uleb128 3                          # DW_AT_name
     .uleb128 8                          # DW_FORM_string
     .uleb128 2                          # DW_AT_location
     .uleb128 23                         # DW_FORM_sec_offset
     .byte 0                             # EOM(1)
     .byte 0                             # EOM(2)
+    .uleb128 3                          # Abbreviation Code
+    .uleb128 52                         # DW_TAG_variable
+    .byte 0                             # DW_CHILDREN_no
+    .uleb128 3                          # DW_AT_name
+    .uleb128 8                          # DW_FORM_string
+    .uleb128 2                          # DW_AT_location
+    .uleb128 34                         # DW_FORM_loclistx
+    .byte 0                             # EOM(1)
+    .byte 0                             # EOM(2)
     .byte 0                             # EOM(3)
 .LAbbrevEnd:
 
     .section .debug_info.dwo, "e", @progbits
 .LCUBegin:
     .long .LCUEnd-.LCUVersion           # Length of Unit
 .LCUVersion:
     .short 5                            # Version
     .byte 5                             # DW_UT_split_compile
     .byte 8                             # Address size
     .long 0                             # Abbrev offset
     .quad 0x1100001122222222            # DWO id
     .uleb128 1                          # Abbrev [1] DW_TAG_compile_unit
     .uleb128 2                          # Abbrev [2] DW_TAG_variable
     .asciz "a"                          # DW_AT_name
-    .long .LLL0-.LLLBegin               # DW_AT_location
+    .long .LLL0-.LLLBegin               # DW_AT_location (DW_FORM_sec_offset)
+    .uleb128 3                          # Abbrev [3] DW_TAG_variable
+    .asciz "b"                          # DW_AT_name
+    .uleb128 1                          # DW_AT_location (DW_FORM_loclistx)
     .byte 0                             # End Of Children Mark
 .LCUEnd:
 
 .section .debug_loclists.dwo, "e", @progbits
 ## Start the section with an unused table to check that the reading offset
 ## of the real table is correctly adjusted.
     .long .LLLDummyEnd-.LLLDummyVersion # Length of Unit
 .LLLDummyVersion:
     .short 5                            # Version
     .byte 8                             # Address size
     .byte 0                             # Segment selector size
     .long 0                             # Offset entry count
     .byte 0                             # DW_LLE_end_of_list
 .LLLDummyEnd:
 
 .LLLBegin:
     .long .LLLEnd-.LLLVersion           # Length of Unit
 .LLLVersion:
     .short 5                            # Version
     .byte 8                             # Address size
     .byte 0                             # Segment selector size
-    .long 0                             # Offset entry count
+    .long 2                             # Offset entry count
+.LLLBase:
+    .long .LLL0-.LLLBase
+    .long .LLL1-.LLLBase
 .LLL0:
     .byte 3                             # DW_LLE_startx_length
     .uleb128 1                          # Index
     .uleb128 0x10                       # Length
     .uleb128 1                          # Loc expr size
     .byte 85                            # DW_OP_reg5
     .byte 0                             # DW_LLE_end_of_list
+.LLL1:
+    .byte 3                             # DW_LLE_startx_length
+    .uleb128 5                          # Index
+    .uleb128 0x20                       # Length
+    .uleb128 2                          # Loc expr size
+    .byte 144                           # DW_OP_regx
+    .uleb128 5                          # RDI
+    .byte 0                             # DW_LLE_end_of_list
 .LLLEnd:
 
     .section .debug_cu_index, "", @progbits
 ## Header:
     .short 5                            # Version
     .space 2                            # Padding
     .long 3                             # Section count
     .long 1                             # Unit count