GNU bug report logs - #56861
[PATCH] test-canonicalize-lgpl fails with musl libc

Previous Next

To reply to this bug, email your comments to 56861 AT debbugs.gnu.org.

Toggle the display of automated, internal messages from the tracker.

Message #5 received at submit <at> debbugs.gnu.org (full text, mbox):

From: Sören Tempel <soeren <at> soeren-tempel.net>
To: bug-sed <at> gnu.org
Subject: [PATCH] test-canonicalize-lgpl fails with musl libc
Date: Sun, 31 Jul 2022 15:44:13 +0200

Hi,

While enabling the sed test suite for the Alpine Linux sed package, I
noticed that the test-canonicalize-lgpl test case fails because musl
does not provide canonicalize_file_name and hence sed uses a compat
implementation which invokes realpath(3). However, musl realpath(3)
implementation does not strip leading // characters hence the following
test from gnulib-tests/test-canonicalize-lgpl.c fails (handling of
leading // is implementation-defined in POSIX):

	203     /* On IBM z/OS, "/" and "//" are distinct, yet they both have
	204        st_dev == st_ino == 1.  */
	205 #ifndef __MVS__
	206     if (SAME_INODE (st1, st2))
	207       {
	208         ASSERT (strcmp (result1, "/") == 0);
	209         ASSERT (strcmp (result2, "/") == 0);
	210       }
	211     else
	212 #endif

The problem being the assertions in line 208 and 209.

The following patch fixes this issue:

diff -upr sed-4.8.orig/gnulib-tests/test-canonicalize-lgpl.c sed-4.8/gnulib-tests/test-canonicalize-lgpl.c
--- sed-4.8.orig/gnulib-tests/test-canonicalize-lgpl.c	2022-07-31 15:28:20.249302470 +0200
+++ sed-4.8/gnulib-tests/test-canonicalize-lgpl.c	2022-07-31 15:30:03.099518653 +0200
@@ -200,9 +200,9 @@ main (void)
     ASSERT (result2);
     ASSERT (stat ("/", &st1) == 0);
     ASSERT (stat ("//", &st2) == 0);
-    /* On IBM z/OS, "/" and "//" are distinct, yet they both have
-       st_dev == st_ino == 1.  */
-#ifndef __MVS__
+    /* On IBM z/OS and musl libc, "/" and "//" are distinct, yet
+       they both have st_dev == st_ino == 1.  */
+#if !defined(__MVS__) && !defined(MUSL_LIBC)
     if (SAME_INODE (st1, st2))
       {
         ASSERT (strcmp (result1, "/") == 0);

Greetings,
Sören

PS: Please CC me, I am not subscribed to the list.

Message #8 received at 56861 <at> debbugs.gnu.org (full text, mbox):

From: Sam James <sam <at> gentoo.org>
To: 56861 <at> debbugs.gnu.org
Subject: Bug 56861
Date: Sat, 31 Dec 2022 22:27:41 +0000

[Message part 1 (text/plain, inline)]

This is a gnulib bug and should be been fixed by:
1. https://git.savannah.gnu.org/cgit/gnulib.git/commit/?id=aae308a3354bb3d0ae0bdf8647ce9be8e46abbce
2. https://git.savannah.gnu.org/cgit/gnulib.git/commit/?id=6ee6c7189f4bafbb603ef043fdf44b84c3f7532a

sed 4.9 is the first release to have these changes.

[signature.asc (application/pgp-signature, attachment)]

Previous Next

GNU bug tracking system
Copyright (C) 1999 Darren O. Benham, 1997,2003 nCipher Corporation Ltd, 1994-97 Ian Jackson.