Re: [Linaro-TCWG-CI] gcc-17-2325-g4302e16581b4: 10 regressions on arm

Project / Subsystem

gcc / gcc-17-2325-g4302e16581b4

Date

2026-07-16

Proposer

Richard Biener <richard.guenther@gmail.com>

Source type

public_inbox

Consensus

Proposed

Sentiment

/10

Technical tradeoffs

  • Enabling trapping math for test coverage conflicts with fast-math optimizations needed for NEON vectorization tests.
  • The interaction between flags leads to unexpected errors in the test suite, indicating a need for better flag management or test case isolation.

All attributes

project
gcc
subsystem
gcc-17-2325-g4302e16581b4
patch_id
discussion_id
CAFiYyc0rt5sPOBEMxamBHeQo4KMxjrGOLY7bSEVO8YEYZJzccQ@mail.gmail.com
source_type
public_inbox
title
Re: [Linaro-TCWG-CI] gcc-17-2325-g4302e16581b4: 10 regressions on arm
headline
gcc-17-2325-g4302e16581b4: 10 regressions on arm
tldr
A commit adding testcases for trapping math with comparisons caused regressions on ARM due to an interaction with -ffast-math.
proposer
Richard Biener <richard.guenther@gmail.com>
consensus
Proposed
outcome
proposed
sentiment_score
technical_tradeoffs
  • Enabling trapping math for test coverage conflicts with fast-math optimizations needed for NEON vectorization tests.
  • The interaction between flags leads to unexpected errors in the test suite, indicating a need for better flag management or test case isolation.
series_id
series_role
standalone
series_parts
[]
tags
  • gcc
  • ARM
  • vectorization
  • testing
  • regression
bugzilla_url
date
2026-07-16T00:00:00.000Z

Re: [Linaro-TCWG-CI] gcc-17-2325-g4302e16581b4: 10 regressions on arm

This thread discusses 10 regressions on ARM architectures caused by commit gcc-17-2325-g4302e16581b4, which introduced test cases for trapping math operations with comparisons. Richard Biener explains that the regressions stem from an interaction between the new test cases using -ftrapping-math and the existing test suite logic, which sets -ffast-math for NEON vectorization support. This combination leads to unexpected behavior, specifically manifesting as ‘excess errors’ in test cases like vect-111.c.