From fca9018ff9212aa1729ed0d827776ff6f85ad693 Mon Sep 17 00:00:00 2001 From: Pierre Tardy Date: Thu, 23 Oct 2025 10:06:59 +0200 Subject: [PATCH] fix panic when rustc tries to reduce intermediate filenames length with multi byte chars MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit The issue cannot be reproduced with the former testcase of creating external crates because rust refuses to use "external crate 28_找出字符串中第一个匹配项的下标" because it is not a valid indentifier (starts with number, and contain non ascii chars) But still using 28_找出字符串中第一个匹配项的下标.rs as a filename is accepted by previous rustc releases So we consider it valid, and add an integration test for it to catch any regression on other code related to non ascii filenames. --- compiler/rustc_session/src/config.rs | 19 ++++++++++++++++--- tests/run-make/lto-long-filenames/rmake.rs | 2 -- tests/run-make/lto-long-filenames_cn/rmake.rs | 19 +++++++++++++++++++ 3 files changed, 35 insertions(+), 5 deletions(-) create mode 100644 tests/run-make/lto-long-filenames_cn/rmake.rs diff --git a/compiler/rustc_session/src/config.rs b/compiler/rustc_session/src/config.rs index 297df7c2c9765..5180162ae47d6 100644 --- a/compiler/rustc_session/src/config.rs +++ b/compiler/rustc_session/src/config.rs @@ -1207,13 +1207,26 @@ fn maybe_strip_file_name(mut path: PathBuf) -> PathBuf { if path.file_name().map_or(0, |name| name.len()) > MAX_FILENAME_LENGTH { let filename = path.file_name().unwrap().to_string_lossy(); let hash_len = 64 / 4; // Hash64 is 64 bits encoded in hex - let stripped_len = filename.len() - MAX_FILENAME_LENGTH + hash_len; + let hyphen_len = 1; // the '-' we insert between hash and suffix + + // number of bytes of suffix we can keep so that "hash-" fits + let allowed_suffix = MAX_FILENAME_LENGTH.saturating_sub(hash_len + hyphen_len); + + // number of bytes to remove from the start + let stripped_bytes = filename.len().saturating_sub(allowed_suffix); + + // Find the next character boundary at or after `stripped_bytes` so slicing is valid. + let split_at = filename + .char_indices() + .find(|(i, _)| *i >= stripped_bytes) + .map(|(i, _)| i) + .unwrap_or(filename.len()); let mut hasher = StableHasher::new(); - filename[..stripped_len].hash(&mut hasher); + filename[..split_at].hash(&mut hasher); let hash = hasher.finish::(); - path.set_file_name(format!("{:x}-{}", hash, &filename[stripped_len..])); + path.set_file_name(format!("{:x}-{}", hash, &filename[split_at..])); } path } diff --git a/tests/run-make/lto-long-filenames/rmake.rs b/tests/run-make/lto-long-filenames/rmake.rs index 9e0ba63e9f5b8..98ce54e5efc82 100644 --- a/tests/run-make/lto-long-filenames/rmake.rs +++ b/tests/run-make/lto-long-filenames/rmake.rs @@ -9,8 +9,6 @@ //@ ignore-cross-compile -use std::fs; - use run_make_support::{rfs, rustc}; // This test make sure we don't get such following error: diff --git a/tests/run-make/lto-long-filenames_cn/rmake.rs b/tests/run-make/lto-long-filenames_cn/rmake.rs new file mode 100644 index 0000000000000..f129b686012dc --- /dev/null +++ b/tests/run-make/lto-long-filenames_cn/rmake.rs @@ -0,0 +1,19 @@ +//@ ignore-cross-compile + +use run_make_support::{rfs, rustc}; + +// This test make sure we don't crash when lto creates output files with long names. +// cn characters can be multi-byte and thus trigger the long filename reduction code more easily. +// we need to make sure that the code is properly generating names at char boundaries. +// as reported in issue #147975 +fn main() { + let lto_flags = ["-Clto", "-Clto=yes", "-Clto=off", "-Clto=thin", "-Clto=fat"]; + for prefix_len in 0..4 { + let prefix: String = std::iter::repeat("_").take(prefix_len).collect(); + let main_file = format!("{}28_找出字符串中第一个匹配项的下标.rs", prefix); + rfs::write(&main_file, "fn main() {}\n"); + for flag in lto_flags { + rustc().input(&main_file).arg(flag).run(); + } + } +}