[PATCH] D95916: DebugInfo/Symbolize: Allow STT_NOTYPE/STT_GNU_IFUNC symbols for .symtab symbolization

[David Blaikie via Phabricator via llvm-commits]

dblaikie added a comment.

In D95916#2538260 <https://reviews.llvm.org/D95916#2538260>, @MaskRay wrote:

> Note: There are some interesting tests which can be made easy by allowing ld.lld, e.g. multiple STT_FILE.

Couldn't that be done with an assembly file with multiple `.file` directives? (as much as possible, it'd be good to keep test coverage in the project where the code is - with the cross-project lit testing area more for end-to-end/extra integration (though targeted/narrowly scoped) coverage)



================
Comment at: llvm/test/DebugInfo/Symbolize/ELF/symtab-notype.s:13-14
+# CHECK-EMPTY:
+# CHECK-NEXT:  ??:0:0
+# CHECK-NEXT:  ??:0:0
+# CHECK-EMPTY:
----------------
what are these two lines of output from? (if the gap is covered by the checks on 18 and 19?


================
Comment at: llvm/test/DebugInfo/Symbolize/ELF/symtab-notype.s:26-34
+# CHECK-NEXT:  l_notype_nosize
+# CHECK-NEXT:  ??:0:0
+# CHECK-EMPTY:
+# CHECK-NEXT:  l_notype_nosize
+# CHECK-NEXT:  ??:0:0
+# CHECK-EMPTY:
+# CHECK-NEXT:  l_notype_nosize
----------------
Is this consistent with gnu addr2line behavior? Seems slightly weird to symbolize addresses outside the size of the section, I'd have thought?


================
Comment at: llvm/test/DebugInfo/Symbolize/ELF/symtab-notype.s:44-45
+g_notype:
+  nop
+  retq
+.size g_notype, . - g_notype
----------------
any reason for the variety of instructions? I'd have thought nops might be the clearest/most uninteresting instructions to use here (& in the other test)


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D95916/new/

https://reviews.llvm.org/D95916