llvm-objdump --adjust-vma does not adjust symbol table

[mysterymath]

The llvm-objdump version of the --adjust-vma flag does not adjust the addresses of symbol table entries, while the binutils objdump does. This seems to have been an oversight, as symbol addresses are elsewhere adjusted (e.g., in the disassembly).

Example from the adjust-vma LLVM test case:

$ objdump --all-headers -D -z --adjust-vma=0x123000 adjust-vma.test.tmp
...
SYMBOL TABLE:
0000000000123001 l     F .text  0000000000000000 func
0000000000123000 l       .text  0000000000000000 sym
0000000000123000 l    d  .text  0000000000000000 .text

$ llvm-objdump --all-headers -D -z --adjust-vma=0x123000 adjust-vma.test.tmp
...
SYMBOL TABLE:
0000000000000001 l     F .text  0000000000000000 func
0000000000000000 l       .text  0000000000000000 sym
0000000000000000 l    d  .text  0000000000000000 .text

[llvmbot]

Hi!

This issue may be a good introductory issue for people new to working on LLVM. If you would like to work on this issue, your first steps are:

1. Assign the issue to you.
2. Fix the issue locally.
3. Run the test suite locally.
   3.1) Remember that the subdirectories under test/ create fine-grained testing targets, so you can
   e.g. use make check-clang-ast to only run Clang's AST tests.
4. Create a git commit
5. Run git clang-format HEAD~1 to format your changes.
6. Submit the patch to Phabricator.
   6.1) Detailed instructions can be found here

For more instructions on how to submit a patch to LLVM, see our documentation.

If you have any further questions about this issue, don't hesitate to ask via a comment on this Github issue.

@llvm/issue-subscribers-good-first-issue

[llvmbot]

@llvm/issue-subscribers-tools-llvm-objdump

[da-viper]

@mysterymath do you have the test file you used for the example or how to create the test file

I tried using the compile output from this

int main (int, char**) { 
    return 0;
}

But the output is the same as not adding the --adjust-vma=0x123000

Thanks

[mysterymath]

@da-viper, I obtained the example by reusing the temporary output of the existing test case:
llvm/test/tools/llvm-objdump/X86/adjust-vma.test

The file can be generated directly by running yaml2obj on the test:
$ <build_dir>/bin/yaml2obj <llvm_dir>/llvm/test/tools/llvm-objdump/X86/adjust-vma.test -o example.elf

[HamidrezaSK]

Hi @mysterymath,

I want to work on this issue. Could you assign it to me?

[HamidrezaSK]

I believe I have resolved the problem, but there is still one remaining issue. When I run the test "ninja check-llvm-tools-llvm-objdump," the adjust-vma test fails, which is expected. Should I modify the test as well? Specifically, should I include the symbol table instance for ADJUST check-prefix?

[mysterymath]

Yeah, modify the test as well. It's definitely strange that this case was tested without the values being adjusted, but both the expected use case of the feature and objdump's behavior suggest that they should be.

[HamidrezaSK]

Hi @mysterymath,

I submitted my patch D153027 for review. The Linux build is OK. However, the Windows build encounters a failure in the "split-dwarf-local-import3.ll" test, and I'm having difficulty understanding why. I would appreciate it if you could provide some guidance in this case.

[mysterymath]

That test is newly added and doesn't invoke llvm-objdump at all; seems pretty safe to call it a flake. :)

[HamidrezaSK]

Hi,

The patch I sent was accepted. Does this imply that the issue is now considered closed?

[EugeneZelenko]

@HamidrezaSK: Issue should be closed after patch committed to main.

[mysterymath]

@HamidrezaSK , I can land the patch for you, but it doesn't cleanly apply to tip of tree since D153401 landed. Can you sync, resolve the merge conflicts, and reupload?