Skip site navigation (1)Skip section navigation (2)
Date:      Tue, 28 Jun 2011 23:28:27 +0200
From:      Niclas Zeising <niclas.zeising@gmail.com>
To:        Eric McCorkle <eric@shadowsun.net>
Cc:        freebsd-ports@freebsd.org
Subject:   -Qunused-arguments and clang
Message-ID:  <4E0A477B.3060008@gmail.com>
In-Reply-To: <4E0A4328.3000103@shadowsun.net>
References:  <4E0679AA.20100@shadowsun.net>	<4E06863F.7010104@FreeBSD.org>	<4E0A2172.7000800@daemonic.se>	<86k4c5epbm.fsf@gmail.com> <4E0A343C.6000008@gmail.com> <4E0A4328.3000103@shadowsun.net>

next in thread | previous in thread | raw e-mail | index | archive | help
[Changing subject since previous was erroneous]

On 2011-06-28 23:10, Eric McCorkle wrote:
> On 6/28/11 4:06 PM, Niclas Zeising wrote:
>> On 2011-06-28 21:19, Pan Tsu wrote:
>>> Niclas Zeising<zeising@daemonic.se>  writes:
>>>
>>>> Sorry for hijacking this thread, and cross posting.
>>>>
>>>> On 2011-06-26 03:07, Gabor PALI wrote:
>>>>>
>>>>>
>>>>>> With Clang, an error occurs in one of the configure scripts, because
>>>>>> Clang warns about unused command-line arguments, and the configure
>>>>>> script assumes that to be a compiler error.  You can deal with
>>>>>> this by
>>>>>> adding -Qunused-parameter to CFLAGS.
>>>>>
>>>>> Thanks for investigating this.
>>>>>
>>>>
>>>> This should probably be made the default, at least for ports when clang
>>>> is compiled, since the output generated when not using
>>>> -Qunused-parameter confuses configure scripts, and stops at least
>>>> FireFox 5 from compiling, that I know of.
>>>
>>> Do you use ccache? Try without.
>>>
>>> For example, the combo confuses libtool
>>>
>>
>> It has nothing to do with cccache. The issue is that clang by default
>> warns when passed flags (-std= -f -W etc.) that's not used during the
>> compilation/linking. This can be silenced with -Qunused-arguments, and
>> if not, it confuses configure scripts that believe this is an error in
>> the code it uses to test for features.
>>
>> Regards!
> 
> Example: In the GHC port, the configure script tries to figure out
> whether the ISA supports PC-relative indexing.  It does this by
> generating a simple program, which contains some asm containing
> PC-relative indexing, compiling it, and seeing if the compiler generates
> a warning.  If clang warns about unused arguments, then the configure
> script assumes (falsely) that the ISA does not support PC-relative
> indexing, and generates bad assembly code, which causes trouble downstream.

Firefox 5 does similar things, but it's looking for pc relative
addressing instead, which is supported on my test system, although
without -Qunused-arguments the configure script thinks otherwise, and
generates bad asm.  I think that option should be added by default,
since this is a big difference from gcc, and bound to cause trouble in
several places.

Regards!
-- 
Niclas



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?4E0A477B.3060008>