Re: [jamboost] -isystem and warnings

From: Rene Rivera (grafik666_at_[hidden])
Date: 2002-10-27 23:26:04

[2002-10-27] David Abrahams wrote:

>Rene Rivera <grafik666_at_[hidden]> writes:
>
>> [2002-10-27] David Abrahams wrote:
>> 
>> >
>> >In some of my Boost.Build v2 testing I've encountered warnings from
>> >GCC because we're normally using -isystem to add Boost headers as part
>> >of the #include <...> -only search path (there appears to be no other
>> >way). I'm beginning to think that we have to give up on -isystem,
>> >since it causes so many problems, including that on some platforms it
>> >adds an implicit extern "C" { } wrapper. For the record, I still think
>> >we're doing the right thing by using angle-includes for boost.
>> 
>> That's fine for GCC, but if we switch to "-I" we'll probably get other
>> warnings. 
>
>I don't understand what you're saying. We'll get warnings with other
>compilers if we switch GCC to use "-I" for <sysinclude>? You can't
>mean _that_.

I do mean it... I used to get some rather silly warnings, like the very
anoying "no newline at eof" but only for <> files not "" files without
-isystem. I remember others but can't rmember the specifics, and it was in
my code not Boost code ;-)

>> Of course the reason it probably causes problems is that it
>> doesn't seem to be a well tested part of GCC :-(
>
>Yeah, it only gets used on ALL THE SYSTEM HEADERS! <wink>

Many instances of one use pattern, does not make many test cases ;-)

>> >Anyway, the warnings I've been seeing are:
>> >
>> > boost/cast.hpp:178: warning: decimal constant is so large that it is
>> unsigned
>> >
>> >This one worries me a little. Has anyone tested the fencepost
>> >condition here to make sure it's doing the right job?
>> 
>> I've seen that one but I was under the impression that it was not related
at
>> all to -isystem. 
>
>I guess I wasn't clear. -isystem suppresses some warnings, including
>the one above. Boost.Build v2 is just using -I so it shows up.
>
>> I've seen that warning on the darwin-tools.jam file I've been trying
>> to get working and it's using -I (as -isystem does the extern
>> wrapping).
>> 
>> Are we getting warnings that go away if we don't use -isystem?
>
>No, the other way around. We're not getting warnings that we'd see if
>we were using -I.

Ahh got it... I'm fine with changing it to "-I". Even though "-isystem" is
the "correct" way, but if it doesn't work, oh well.

-- grafik - Don't Assume Anything
-- rrivera_at_[hidden] - grafik_at_[hidden]
-- 102708583_at_icq - Grafik666_at_AIM - Grafik_at_[hidden]