Re: [PATCH] RISC-V: Also output undefined symbols as dynamic symbols

← All discussions

Project / Subsystem

binutils / risc-v

Date

2026-05-13

Proposer

Nelson Chu <nelson.chu1990@gmail.com>

Source type

public_inbox

Consensus

Proposed

Sentiment

—/10

Technical tradeoffs

• Fixes a regression in RISC-V linking behavior.
• Ensures correct handling of undefined symbols during dynamic linking.
• The change may be specific to RISC-V due to its unique linking requirements.

View thread →

All attributes

project: binutils
subsystem: risc-v
patch_id: —
discussion_id: CAHT-pJzqthybcirJLPU0Rus_5pOmpOnUVgqfpf5Cr7VHLnza3g@mail.gmail.com
source_type: public_inbox
title: Re: [PATCH] RISC-V: Also output undefined symbols as dynamic symbols
headline: RISC-V: Also output undefined symbols as dynamic symbols
tldr: A RISC-V linker change ensures undefined symbols are treated as dynamic symbols when linking with `--unresolved-symbols=ignore-in-object-file`, fixing crashes…
proposer: Nelson Chu <nelson.chu1990@gmail.com>
consensus: Proposed
outcome: proposed
sentiment_score: —
technical_tradeoffs: • Fixes a regression in RISC-V linking behavior.
• Ensures correct handling of undefined symbols during dynamic linking.
• The change may be specific to RISC-V due to its unique linking requirements.
series_id: —
series_role: standalone
series_parts: []
tags: • binutils
• risc-v
• linker
• dynamic linking
• undefined symbols
url: https://inbox.sourceware.org/binutils/CAHT-pJzqthybcirJLPU0Rus_5pOmpOnUVgqfpf5Cr7VHLnza3g@mail.gmail.com
bugzilla_url: —
date: 2026-05-13T00:00:00.000Z

Re: [PATCH] RISC-V: Also output undefined symbols as dynamic symbols

Vivian Wang identified that recent binutils changes cause RISC-V linkers to misbehave when encountering undefined symbols, specifically crashing or throwing errors when the desired behavior (with --unresolved-symbols=ignore-in-object-file) is to leave the symbol unresolved for the dynamic linker. The proposed patch restores the emission of dynamic symbols for undefined symbols, fixing a regression and a failing test case. Nelson Chu questions the RISC-V-specific nature of the fix.