Re: [PATCH] elf: Hide DT_INIT and DT_FINI functions
Project / Subsystem
binutils / binutils
Date
2026-09-15
Proposer
Alan Modra <amodra@gmail.com>
Source type
public_inbox
Consensus
Proposed
Sentiment
—/10
Technical tradeoffs
- • Hiding linker-internal symbols vs. providing source-level control for initialization/termination functions.
- • Potential impact on dynamic linking behavior and runtime initialization/cleanup compared to the complexity of linker-level hiding.
All attributes
- project
- binutils
- subsystem
- binutils
- patch_id
- —
- discussion_id
- aqim7jougG6iio7x@squeak.grove.modra.org
- source_type
- public_inbox
- title
- Re: [PATCH] elf: Hide DT_INIT and DT_FINI functions
- headline
- Reverted patch to hide DT_INIT and DT_FINI functions
- tldr
- Patch to hide DT_INIT/DT_FINI functions in ELF reverted due to negative community feedback.
- proposer
- Alan Modra <amodra@gmail.com>
- consensus
- Proposed
- outcome
- needs_revision
- sentiment_score
- —
- technical_tradeoffs
-
- • Hiding linker-internal symbols vs. providing source-level control for initialization/termination functions.
- • Potential impact on dynamic linking behavior and runtime initialization/cleanup compared to the complexity of linker-level hiding.
- series_id
- —
- series_role
- standalone
- series_parts
- []
- tags
-
- • binutils
- • elf
- • linker
- • abi
- • reverted
- bugzilla_url
- —
- date
- 2026-09-15T00:00:00.000Z
Re: [PATCH] elf: Hide DT_INIT and DT_FINI functions
Alan Modra has reverted a patch that aimed to hide DT_INIT and DT_FINI functions in ELF objects. The decision was made because the majority opinion was that ld hiding these functions is surprising, they can be hidden at the source level, and the impact of them not being hidden is small. There might even be reasons to keep them exported. The patch also introduced regressions that required further fixes.