Re: [PATCH v2] kconfig/kbuild: define _POSIX_C_SOURCE

[Arnaud Lacombe]

Hi,

On Tue, Apr 19, 2011 at 5:54 AM, Michal Marek <mmarek@xxxxxxx> wrote:
> On 10.4.2011 16:31, Valentin Ochs wrote:
>> On Sat, Apr 09, 2011 at 11:34:00PM -0400, Arnaud Lacombe wrote:
>>> On Sat, Apr 9, 2011 at 10:09 PM, Valentin Ochs <a@xxxxxx> wrote:
>>>> The three patched files use PATH_MAX without defining the required
>>>> _POSIX_C_SOURCE feature test macro.  This prevents compilation with the
>>>> musl libc.  The patch applies to 2.6.38.2.
>>>>
>>>> Changes since v1:
>>>>  - fix scripts/kconfig/lex.zconf.c_shipped
>>> this file is autogenerated from zconf.l, which should be updated as
>>> well. I'm not sure how you searched for PATH_MAX, but you're still
>>> missing `confdata.c' and `nconf.c'.
>
> Yes, please fix the bison parser and run make GENERATE_PARSER=1
> menuconfig to update the *_shipped files.
>
>
>>> I had a quick look to different libc implementation, glibc and uClibc
>>> default to _POSIX_C_SOURCE == 200112L, FreeBSD >8.0 defaults to
>>> 200809L for >8.0, FreeBSD 7.x to 200112L.
>>
>> I got the value I used from the Open Group System Interfaces
>> specification, but the musl author said that 200112L would be fine too.
>
> I suggest you use 200112L, so that it is a nop for glibc/uClibc builds.
> Please also accompany it with a /* for PATH_MAX */ comment or so.
>
no it's not. Defining _POSIX_C_SOURCE disables automatic definition of
a few features macros, see below.

>>
>>> None of these seems to requires _POSIX_C_SOURCE to define PATH_MAX, so
>>> I'm not certain of the requirement of the change.
>>
>> While I don't want to appear like a language lawyer, the specification
>> says that 'all symbols required by POSIX.1-2008 to appear when the
>> header is included shall be made visible' when an application defines
>> _POSIX_C_SOURCE. I guess the musl author interprets that as 'if you
>> don't define the feature test macros, you're not getting PATH_MAX.' This
>> does not seem to be incorrect behaviour to me.
>
> While I don't completely understand the motivation for such
> super-pedantic libc implementation, the fact is that POSIX says one
> should define this macro, so let's define it, it does not hurt us.
>
If we go that way, we should at least:

diff --git a/Makefile b/Makefile
index d6592b6..0490413 100644
--- a/Makefile
+++ b/Makefile
@@ -233,7 +233,7 @@ CONFIG_SHELL := $(shell if [ -x "$$BASH" ]; then
echo $$BASH; \

 HOSTCC       = gcc
 HOSTCXX      = g++
-HOSTCFLAGS   = -Wall -Wmissing-prototypes -Wstrict-prototypes -O2
-fomit-frame-pointer
+HOSTCFLAGS   = -D_POSIX_C_SOURCE=200112L -Wall -Wmissing-prototypes
-Wstrict-prototypes -O2 -fomit-frame-pointer
 HOSTCXXFLAGS = -O2

at best:

index d6592b6..0490413 100644
--- a/Makefile
+++ b/Makefile
@@ -233,7 +233,7 @@ CONFIG_SHELL := $(shell if [ -x "$$BASH" ]; then
echo $$BASH; \

 HOSTCC       = gcc
 HOSTCXX      = g++
-HOSTCFLAGS   = -Wall -Wmissing-prototypes -Wstrict-prototypes -O2
-fomit-frame-pointer
+HOSTCFLAGS   = --std=c99 -D_POSIX_C_SOURCE=200112L -Wall
-Wmissing-prototypes -Wstrict-prototypes -O2 -fomit-frame-pointer
 HOSTCXXFLAGS = -O2

Both introduces new warnings with the glibc shipped with fedora 14
relating to strdup(3) needing one of: _SVID_SOURCE, _BSD_SOURCE,
_XOPEN_SOURCE, _POSIX_C_SOURCE>=200809L (for __USE_XOPEN2K8).
-D_POSIX_C_SOURCE=200809L do not generate more warning, at least on my
host.

Just for reference, the second change fails anyway as `nconf' uses GNU
extension of C99.

 - Arnaud

> Michal
> --
> To unsubscribe from this list: send the line "unsubscribe linux-kbuild" in
> the body of a message to majordomo@xxxxxxxxxxxxxxx
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
>
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/