civodul pushed a commit to branch main
in repository shepherd.
commit d98d61a8a3f20de46d18ce4a8af05c93fab20b89
Author: Ludovic Courtès <[email protected]>
AuthorDate: Mon Mar 31 23:04:52 2025 +0200
system-log: Work around ‘regexp-exec’ bug in non-Unicode capable locale.
Fixes <https://issues.guix.gnu.org/77283>.
* modules/shepherd/service/system-log.scm (%system-log-message-rx):
Remove suffix from regexp.
(parse-system-log-message): Call ‘match:suffix’.
* tests/services/system-log-internal.scm (call-with-c-locale): New
procedure.
(with-c-locale): New macro.
("read-system-log-message, with Unicode and C locale"): New test.
* NEWS: Update.
Reported-by: MSavoritias <[email protected]>
---
NEWS | 8 ++++++++
modules/shepherd/service/system-log.scm | 10 ++++++----
tests/services/system-log-internal.scm | 27 ++++++++++++++++++++++++++-
3 files changed, 40 insertions(+), 5 deletions(-)
diff --git a/NEWS b/NEWS
index e6a7cfe..b6663e3 100644
--- a/NEWS
+++ b/NEWS
@@ -61,6 +61,14 @@ would be incorrectly calculated when the event would fall
between 02:00am and
03:00am, leading the timer to trigger many times in a row, unless it had
#:wait-for-termination? #true. This is now fixed; next year will be better!
+** System log no longer crashes on some Unicode input
+ (<https://issues.guix.gnu.org/77283>)
+
+Due to a bug in Guile’s (ice-9 regex) module, the ‘system-log’ service could
+crash on certain inputs containing non-ASCII Unicode characters, when
+‘shepherd’ is running in a non-Unicode capable locale (which is usually the
+case for PID 1). This is now fixed by sidestepping the regexp bug entirely.
+
* Changes in 1.0.3
** ‘spawn-command’ now honors #:log-file
diff --git a/modules/shepherd/service/system-log.scm
b/modules/shepherd/service/system-log.scm
index bf9ca69..fb31c12 100644
--- a/modules/shepherd/service/system-log.scm
+++ b/modules/shepherd/service/system-log.scm
@@ -122,10 +122,12 @@
%system-log-priority-mask))
(define %system-log-message-rx
- ;; Regexp matching system log messages. Example:
+ ;; Regexp matching the prefix of system log messages. Example:
;; <29>Jun 22 16:41:30 wpa_supplicant[303]: whatever
- (make-regexp "<([0-9]+)> ?([[:alpha:]]{3} +[0-9]+ [0-9]+:[0-9]+:[0-9]+ )?\
-(.*)"))
+ ;; Note: Do not match the actual message (the suffix) because it might
+ ;; contain Unicode code points and 'regexp-exec' might be buggy in that
+ ;; case: <https://bugs.gnu.org/77392>.
+ (make-regexp "<([0-9]+)> ?([[:alpha:]]{3} +[0-9]+ [0-9]+:[0-9]+:[0-9]+ )?"))
(define %default-priority (system-log-priority notice))
(define %default-facility (system-log-facility user))
@@ -140,7 +142,7 @@ representing it."
(m
(let* ((facility+priority (string->number (match:substring m 1))))
(system-log-message facility+priority
- (match:substring m 3) sender)))))
+ (match:suffix m) sender)))))
(define* (read-system-log-message port #:optional sender)
"Read a system log message from @var{port}. Return the end-of-file object
diff --git a/tests/services/system-log-internal.scm
b/tests/services/system-log-internal.scm
index 87e36c1..d4280c1 100644
--- a/tests/services/system-log-internal.scm
+++ b/tests/services/system-log-internal.scm
@@ -1,5 +1,5 @@
;; GNU Shepherd --- Test the system log service.
-;; Copyright © 2024 Ludovic Courtès <[email protected]>
+;; Copyright © 2024-2025 Ludovic Courtès <[email protected]>
;;
;; This file is part of the GNU Shepherd.
;;
@@ -20,6 +20,17 @@
#:use-module (shepherd service system-log)
#:use-module (srfi srfi-64))
+(define (call-with-c-locale thunk)
+ (let ((previous #f))
+ (dynamic-wind
+ (lambda () (set! previous (setlocale LC_ALL "C")))
+ thunk
+ (lambda () (setlocale LC_ALL previous)))))
+
+(define-syntax-rule (with-c-locale exp)
+ (call-with-c-locale (lambda () exp)))
+
+
(test-begin "system-log-internal")
(test-equal "read-system-log-message, with PID"
@@ -46,6 +57,20 @@ sudo: ludo : TTY=pts/0 ; PWD=/home/ludo ; USER=root ;
COMMAND=xyz"
(system-log-message-priority message)
(system-log-message-content message))))))
+(test-equal "read-system-log-message, with Unicode and C locale"
+ (list (system-log-facility daemon)
+ (system-log-priority debug)
+ "prosody[198]: Ϫ")
+ ;; This used to trigger a bug whereby 'regexp-exec' (in Guile 3.0.9) would
+ ;; incorrectly compute match boundaries when running in a non-Unicode
+ ;; capable locale: <https://issues.guix.gnu.org/77283>.
+ (call-with-input-string "<31>Mar 26 16:41:31 prosody[198]: Ϫ"
+ (lambda (port)
+ (let ((message (with-c-locale (read-system-log-message port))))
+ (list (system-log-message-facility message)
+ (system-log-message-priority message)
+ (system-log-message-content message))))))
+
(test-equal "read-system-log-message, raw"
(list (system-log-facility user)
(system-log-priority notice)
