Hi Ben,
> I noticed that (sometimes?) the lseek check prints a "broken
> pipe" message in the middle of "configure" output, e.g.
> checking whether lseek detects pipes... ../configure: line 30326:
> echo: write error: Broken pipe
> as reported here under step 19:
> http://benpfaff.org/~blp/pspp-master/20100220040512/
Below is a proposed fix (untested).
> I think that we should redirect stderr to /dev/null for that
> test.
I would not do this. "2> /dev/null" is a recipe for creating buggy software:
When an error creeps in, for whatever reason, you won't notice it.
Bruno
2010-02-20 Bruno Haible <br...@clisp.org>
Silence a "broken pipe" message from bash.
* m4/lseek.m4 (gl_FUNC_LSEEK): Ignore SIGPIPE in a subcommand.
Reported by Ben Pfaff <b...@cs.stanford.edu>.
--- m4/lseek.m4.orig Sat Feb 20 21:18:21 2010
+++ m4/lseek.m4 Sat Feb 20 21:17:50 2010
@@ -1,4 +1,4 @@
-# lseek.m4 serial 4
+# lseek.m4 serial 5
dnl Copyright (C) 2007, 2009, 2010 Free Software Foundation, Inc.
dnl This file is free software; the Free Software Foundation
dnl gives unlimited permission to copy and/or distribute it,
@@ -21,7 +21,12 @@
}],
[if test -s conftest$ac_exeext \
&& ./conftest$ac_exeext < conftest.$ac_ext \
- && { echo hi | ./conftest$ac_exeext; test $? = 1; }; then
+ && { : "Avoid error message from bash."; \
+ : "See func_reset_sigpipe in gnulib-tool."; \
+ if test -n "$BASH_VERSION"; then trap - SIGPIPE; fi; \
+ echo hi | ./conftest$ac_exeext; \
+ test $? = 1; \
+ }; then
gl_cv_func_lseek_pipe=yes
else
gl_cv_func_lseek_pipe=no
