[cfe-dev] Patch to quiet some options

[Daniel Dunbar]

On Tue, Sep 1, 2009 at 5:59 AM, David Chisnall<csdavec at swansea.ac.uk> wrote:
> On 1 Sep 2009, at 06:54, Daniel Dunbar wrote:
>
>> On Mon, Aug 31, 2009 at 4:01 PM, David Chisnall<csdavec at swansea.ac.uk>
>> wrote:
>>>
>>> On 18 Aug 2009, at 06:59, Daniel Dunbar wrote:
>>>
>>>> I think the right fix is to pass -pthread to clang-cc since it may
>>>> change the language (depending on the target?). This will have the
>>>> side effect of silencing the warning.
>>>
>>> The attached diff passes -pthread on to clang-cc, which then ignores it.
>>>  When someone comes to add support for platforms like HP-UX where it
>>> actually does something there is now a place for them to do so.  On
>>> platforms where it is silently ignored by GCC (I think that includes
>>> everything we support so far), it now doesn't issue an unrecognised
>>> argument
>>> warning.
>>
>> Ok, but, might as well add the language option and preprocessor define
>> while we are at it?
>
> Well, the problem with this is that I have no idea what they are because I
> don't seem to have access to any platforms where -pthread does anything
> other than imply -lpthread while linking.  Apparently on some platforms it
> defines a preamble to the executable, on some it defines macros (but I don't
> know what they are) and (I think) it also alters the behaviour of OpenMP on
> some platforms (but we don't yet support OpenMP), so it seemed more sensible
> to just leave the option there and let people define sensible language
> options when they add a target or feature that uses this flag, rather than
> for me to add some based on guesses that would sit unused and then be
> replaced later.

This is more or less why I like the warning; its a red flag to
indicate something we haven't implemented yet. If you file a bugzilla
on this I'll add it when I get home.

 - Daniel