fix(debuginfo): name size-0 asm labels instead of raw addresses

[not-matthias]

TLDR: Hand-written ASM stubs symbols' were not resolved. This PR fixes it

Hand-written assembly entry points often carry no .size directive, so they
appear as size-0 STT_NOTYPE symtab labels with no covering DWARF
subprogram. get_elf_symbol_info rejected them twice (STT_NOTYPE isn't
plausible, and size-0 symbols are dropped outside Android), leaving those
addresses unnamed. Callgrind then falls back to printing the raw address,
which is how the root frame of exec'd CLI benchmarks turned from _dl_init
into 0x000000000001f59f: the true caller is glibc's _dl_start_user, a
size-0 label in ld.so.

Before:

After:

The ASM func from the screenshots: https://github.com/bminor/glibc/blob/04e750e75b73957cf1c791535a3f4319534a52fc/sysdeps/or1k/dl-start.S#L53

[not-matthias]

@greptileai review

[greptile-apps]

Greptile Summary

The PR recognizes non-empty, size-zero STT_NOTYPE labels in executable sections and gives them synthetic ranges while preserving precedence for real symbols.

• Propagates synthetic-label metadata through ELF symbol loading.
• Drops synthetic labels covered by real symbols and trims remaining overlaps during canonicalization.
• Adds an amd64 assembly regression test for standalone and interior labels.

Confidence Score: 4/5

The PR is not yet safe to merge because an oversized executable section can still make synthetic-label sizing abort debug-information loading.

The current assignment narrows an address-sized section remainder into a signed symbol size without enforcing the existing signed-size bound, and that value is subsequently used to construct a range whose ordering is asserted.

Files Needing Attention: coregrind/m_debuginfo/readelf.c

Important Files Changed

Filename	Overview
coregrind/m_debuginfo/readelf.c	Admits size-zero text labels and assigns synthetic sizes, but the previously reported unchecked narrowing into Int remains.
coregrind/m_debuginfo/storage.c	Prioritizes real symbols, removes covered synthetic labels, and preserves synthetic metadata during overlap resolution.
coregrind/m_debuginfo/priv_storage.h	Adds the isSynthLabel marker to DiSym.
none/tests/amd64/synth_label_asm.S	Defines standalone and interior size-zero assembly labels for regression coverage.
none/tests/amd64/synth_label.c	Exercises backtrace symbolization through synthetic and conventionally sized assembly entries.
none/tests/amd64/Makefile.am	Registers the new amd64 regression test, its sources, and filtering assets.

Flowchart

%%{init: {'theme': 'neutral'}}%%
flowchart TD
    A[Read ELF symbol] --> B{Size-zero STT_NOTYPE label\nin text section?}
    B -- No --> C[Existing symbol handling]
    B -- Yes --> D[Assign synthetic remaining-section size]
    D --> E[Add symbol with isSynthLabel]
    E --> F[Sort symbol table]
    F --> G{Starts inside real symbol?}
    G -- Yes --> H[Drop synthetic label]
    G -- No --> I[Trim against successor]
    I --> J[Use label for symbolization]

Loading

_{Reviews (2): Last reviewed commit: "fix(debuginfo): name size-0 asm labels i..." | Re-trigger Greptile}

[codspeed-hq]

Merging this PR will not alter performance

✅ 84 untouched benchmarks
⏩ 60 skipped benchmarks¹

---

_{Comparing cod-3270-exec-cli-benchmarks-in-e2e-tests-fail-to-parse-_dl_init (4d00230) with master (ae6bf15)}

Footnotes

1. 60 benchmarks were skipped, so the baseline results were used instead. If they were deleted from the codebase, click here and archive them to remove them from the performance reports. ↩

[GuillaumeLagrange]

If it's not possible, then I'll approve the current implem