Landing: 63851c77785a

← All landings

Project / Subsystem

gcc / gcc/libgomp

Date

2026-06-12

Author

Lipeng Zhu

Commit

63851c77785af1e51c56a5614d324a02430de382

Source

github

Perf win

Breaking

View on GitHub →

All attributes

project: gcc
subsystem: gcc/libgomp
patch_id: —
commit_hash: 63851c77785af1e51c56a5614d324a02430de382
source_type: github
headline: Libgomp test ensures unique filenames with many threads
tldr: A libgomp Fortran test now uses a larger buffer to correctly handle I/O unit numbers with more digits, preventing filename collisions in multi-threaded environ…
author: Lipeng Zhu
outcome: committed
performance_win: false
breaking_change: false
series_id: —
series_parts: []
tags: • libgomp
• fortran
• testing
• concurrency
discussion_id_link: —
bugzilla_pr: —
url: https://github.com/gcc-mirror/gcc/commit/63851c77785af1e51c56a5614d324a02430de382
date: 2026-06-12T00:00:00.000Z

The libgomp.fortran/rwlock_1.f90 test in GCC’s OpenMP library now uses an enlarged buffer for file_name to accommodate 5-digit I/O unit numbers, up from 3 digits. This change addresses a potential issue where, with many threads, the generated temporary filenames for I/O operations could collide due to truncated unit numbers, leading to test failures or incorrect behavior in libgomp environments.