fix: explicitly remap current dir by using .

[weihanglo]

What does this PR try to resolve?

In https://github.com/rust-lang/rust/blob/87e1447aa/compiler/rustc_codegen_llvm/src/debuginfo/metadata.rs#L856
when the remap result of work_dir is an empty string,
LLVM won't generate some symbols for root debuginfo node.
For example, N_SO and N_OSO on macOS,
or DW_AT_comp_dir and DW_AT_GNU_dwo_name when debuginfo is splitted.
Precisely, it is observed that when the DIFile of a root DI node of a
compile unit DIFile was provied with an empty Directory string,
LLVM would not emit those symbols for the root DI node.

This behavior is not desired,
resulting in corrupted debuginfo and degrading debugging experience.

This is might not be a bug of --remap-path-prefix in rustc,
since -fdebug-prefix-map in clang 16 could have the same result
(DW_AT_comp_dir is gone when work_dir is remapped to an empty string).
However, in gcc 12 fdebug-prefix-map will return an absolute work_dir
when an empty string occurs.

To not bother whether this needs to be fixed in rustc or not,
let's fix it by always appending an explicit .
when --remap-path-prefix remaps to relative workspace root
a.k.a. where rustc is invoking.

How should we test and review this PR?

Build rustc

Without a possible fix in rustc like rust-lang/rust#118518, the debuginfo is not corrputed even we dont remap to . explicitly. Therefore, to test it manually you need to build rustc from rust-lang/rust#118518. Generally,

# in rust-lang/rust ./x b library --stage 1 rustup toolchain link stage1 build/host/stage1 

Then cargo +stage1 build is available, and you can build package via

CARGO_PROFILE_DEV_TRIM_PATHS=object cargo +stage1 -Ztrim-paths b 

After build a package, I would recommend playing with debuggers, either gdb or lldb.
After remapping, supposedly you need to launch the debugger from the same location the build kicked off, usually the workspace root.

Please help verify that source files still show, and breakpoints can pause/resume at either local or registry packages.

I don't want to build rustc

You can use either objdump -Wi or readelf -wi or dsymutil -s to inspect debug symbols, though it doesn't help much.

Additional information

It would be awesome if somebody can help verify the behavior on Windows 😆.

[rustbot]

r? @ehuss

(rustbot has picked a reviewer for you, use r? to override)

[bors]

☔ The latest upstream changes (presumably #13118) made this pull request unmergeable. Please resolve the merge conflicts.

[weihanglo]

Actually I am not sure if we want to fix this in cargo or in rustc…

[weihanglo]

• Run clang main.c -gdwarf -fdebug-prefix-map="$PWD"= -gsplit-dwarf
  • on Linux DW_AT_comp_dir is gone, as same as the result from rustc.
  • on macOS N_SO and N_OSO are gone, as same as the result from rustc.
• Run gcc main.c -gdwarf -fdebug-prefix-map="$PWD"= -gsplit-dwarf
  • on Linux it generates empty DW_AT_comp_dir.
  • on macOS it generates N_SO with value /.

Without bothering whether this should be fixed in rustc or cargo, I would propose we add an . for remapping the current working directory.

For more on gcc/clang remap options, see: https://reproducible-builds.org/docs/build-path/

[weihanglo]

This is ready for review, but somehow a bit convoluted.

[epage]

@bors r+

[bors]

📌 Commit fcd4221 has been approved by epage

It is now in the queue for this repository.

[bors]

⌛ Testing commit fcd4221 with merge 6feabf9...

[bors]

☀️ Test successful - checks-actions
Approved by: epage
Pushing 6feabf9 to master...