On 12/04/2013 11:51 AM, Paul Eggert wrote: > or we can change the test case to not tickle the bug.
I did the latter, as follows. Maybe at some point this should be filed as a bug against glibc, since the bug occurs even with _FORTIFY_SOURCE defined to 1 and that isn't supposed to break conforming programs. --- ChangeLog | 9 +++++++++ tests/test-open.h | 11 ++++++++++- 2 files changed, 19 insertions(+), 1 deletion(-) diff --git a/ChangeLog b/ChangeLog index cdf69fa..1f2b13c 100644 --- a/ChangeLog +++ b/ChangeLog @@ -1,3 +1,12 @@ +2013-12-05 Paul Eggert <egg...@cs.ucla.edu> + + open-tests: port to glibc with _FORTIFY_SOURCE and -O1 + Problem reported by Daiki Ueno in: + http://lists.gnu.org/archive/html/bug-gnulib/2013-06/msg00052.html + * tests/test-open.h (__always_inline): + New macro, if not already defined. + (test_open): Use it. + 2013-12-04 Eric Blake <ebl...@redhat.com> include_next: minimize code duplication diff --git a/tests/test-open.h b/tests/test-open.h index babb239..ff44b3e 100644 --- a/tests/test-open.h +++ b/tests/test-open.h @@ -16,13 +16,22 @@ /* Written by Bruno Haible <br...@clisp.org>, 2007. */ +/* Make test_open always inline if we're using Fortify, which defines + __always_inline to do that. Do nothing otherwise. This works + around a glibc bug whereby 'open' cannot be used as a function + pointer when _FORTIFY_SOURCE is positive. */ + +#ifndef __always_inline +#define __always_inline +#endif + /* This file is designed to test both open(n,buf[,mode]) and openat(AT_FDCWD,n,buf[,mode]). FUNC is the function to test. Assumes that BASE and ASSERT are already defined, and that appropriate headers are already included. If PRINT, warn before skipping symlink tests with status 77. */ -static int +static int __always_inline test_open (int (*func) (char const *, int, ...), bool print) { int fd; -- 1.8.3.1
