remove some assumptions about dynobj

Project / Subsystem

binutils / binutils

Date

2026-08-30

Proposer

Alan Modra <amodra@gmail.com>

Source type

public_inbox

Consensus

Proposed

Sentiment

—/10

Technical tradeoffs

  • • The changes might require re-evaluation of assumptions in other related linker scripts or backend code.
  • • Ensuring comprehensive testing across all affected architectures is crucial.

All attributes

project
binutils
subsystem
binutils
patch_id
—
discussion_id
apS6I5hd4Fnz7pQU@squeak.grove.modra.org
source_type
public_inbox
title
remove some assumptions about dynobj
headline
binutils: remove assumptions about dynobj
tldr
A patch series cleans up assumptions in various backends that dynamic sections must exist if dynobj is non-NULL.
proposer
Alan Modra <amodra@gmail.com>
consensus
Proposed
outcome
proposed
sentiment_score
—
technical_tradeoffs
  • • The changes might require re-evaluation of assumptions in other related linker scripts or backend code.
  • • Ensuring comprehensive testing across all affected architectures is crucial.
series_id
—
series_role
standalone
series_parts
[]
tags
  • • binutils
  • • linker
  • • elf
  • • backend
  • • patch
bugzilla_url
—
date
2026-08-30T00:00:00.000Z

remove some assumptions about dynobj

This patch series addresses an issue where several BFD backends incorrectly assume that the existence of dynobj implies that dynamic sections have been created. This assumption can fail if dynobj is set without create_dynamic_sections being called, such as when dynobj is assigned a stub BFD early in the linking process. The changes adjust assertions and checks in multiple backends to correctly test for dynamic_sections_created or the existence of specific dynamic section symbols.