Eric Blake wrote: > On 02/19/2012 10:45 AM, Jim Meyering wrote: >> Jérémy Compostella wrote: >> >>>> Thank you for the report and the patch. >>>> Please tell us what type/version of compiler/system you're using. >>>> It's good to include that information, especially since ssize_t >>>> is supposed to be defined via stdio.h. Then it is clear that this >>>> change is working around a problem with non-conforming systems. >>> Here the requested information: >>> - linux kernel: Linux Apollo 3.2.0-rc7+ #75 SMP Thu Jan 5 18:14:54 CET >>> 2012 x86_64 GNU/Linux >>> - gcc (Debian 4.4.5-8) 4.4.5 >>> - libc6-dev 2.11.2-10 >> >> Thanks. >> I've adjusted your commit log to explain why: >> >>>From 933989eca67bd82f011595937002cc284c98772d Mon Sep 17 00:00:00 2001 >> From: =?UTF-8?q?J=C3=A9r=C3=A9my=20Compostella?= >> <jeremy.composte...@gmail.com> >> Date: Sun, 19 Feb 2012 15:38:15 +0100 >> Subject: [PATCH] tests: ls/getxattr-speedup: include <sys/types.h> >> for older headers >> >> Modern <stdio.h> provides ssize_t, but the one from Debian's >> libc6-dev 2.11.2-10 apparently does not. >> * tests/ls/getxattr-speedup: Also include <sys/types.h>. > > Shouldn't we be fixing this in gnulib, rather than making every client > package (including coreutils) work around it?
Is it even a problem in gnulib? Notice that that test is unusual in that it uses $CC without the usual -I options and on code that does not include <config.h>.
