Thanks, r141268.<br><br><div class="gmail_quote">On Wed, Oct 5, 2011 at 1:29 PM, Eli Friedman <span dir="ltr"><<a href="mailto:eli.friedman@gmail.com">eli.friedman@gmail.com</a>></span> wrote:<br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex;">
<div class="im">On Wed, Oct 5, 2011 at 10:43 AM, Ivan Krasin <<a href="mailto:krasin@google.com">krasin@google.com</a>> wrote:<br>
> Eli,<br>
> I have updated the patch. Now I use language-extension-token parameter to<br>
> suppress the warning "extension used".<br>
> Is it fine now? Alternatively, I can introduce more fine-grained warning<br>
> asm-token with higher priority than ext_token_used.<br>
> So far, ext_token_used complains about the following tokens (see<br>
> lib/Basic/IdentifierTable.cpp):<br>
><br>
> else if (LangOpts.GNUKeywords && (Flags & KEYGNU)) AddResult = 1;<br>
> else if (LangOpts.MicrosoftExt && (Flags & KEYMS)) AddResult = 1;<br>
> else if (LangOpts.Borland && (Flags & KEYBORLAND)) AddResult = 1;<br>
> So, it's GNU, Microsoft and Borland extensions.<br>
<br>
</div>-Wlanguage-extension-token is fine.<br>
<span class="HOEnZb"><font color="#888888"><br>
-Eli<br>
</font></span><div class="HOEnZb"><div class="h5"><br>
> On Tue, Oct 4, 2011 at 3:33 PM, Ivan Krasin <<a href="mailto:krasin@google.com">krasin@google.com</a>> wrote:<br>
>><br>
>> like, in my particular case, this warning complains about asm directive,<br>
>> which is not a part of ANSI C (-pedantic)<br>
>><br>
>> On Tue, Oct 4, 2011 at 3:32 PM, Ivan Krasin <<a href="mailto:krasin@google.com">krasin@google.com</a>> wrote:<br>
>>><br>
>>><br>
>>> On Tue, Oct 4, 2011 at 3:24 PM, Eli Friedman <<a href="mailto:eli.friedman@gmail.com">eli.friedman@gmail.com</a>><br>
>>> wrote:<br>
>>>><br>
>>>> On Tue, Oct 4, 2011 at 11:13 AM, Ivan Krasin <<a href="mailto:krasin@chromium.org">krasin@chromium.org</a>><br>
>>>> wrote:<br>
>>>> > Hi llvm team!<br>
>>>> ><br>
>>>> > This patch adds support of -Wno-extension-used to make it possible to<br>
>>>> > suppress the warning:<br>
>>>> ><br>
>>>> > lala.c:1:21: warning: extension used [-pedantic]<br>
>>>> > double lala(double) asm("llvm.log.f64");<br>
>>>> ><br>
>>>> > r140770 has started to promote this warning to error if -Werror is<br>
>>>> > specified (which is absolutely correct).<br>
>>>> > We have a test that checks some llvm intrinsics and it's not Clangy to<br>
>>>> > drop -Werror just because one particular warning is expected.<br>
>>>> > I have added the option to suppress this warning and has removed it<br>
>>>> > from the test for warnings w/o suppression options.<br>
>>>> ><br>
>>>> > OK to commit?<br>
>>>><br>
>>>> We really ought to clarify what exactly this is warning about and name<br>
>>>> the warning group in a more intuitive way. -Wno-extension-used<br>
>>>> doesn't give the user any idea what warning it is actually<br>
>>>> suppressing. I'm drawing a blank as to what exactly to call it,<br>
>>>> though.<br>
>>><br>
>>> Hi Eli,<br>
>>> It suppresses any language extension used in the program source. I have<br>
>>> no preference on how the flag should be named and I have chosen<br>
>>> "extension-used" just to fit to the existing warning message:<br>
>>> "warning: extension used [-pedantic]"<br>
>>> Probably, a better name would be language-extension or<br>
>>> language-extension-used. In this case -Wno-language-extension would give a<br>
>>> good idea of what warning does it suppress.<br>
>>> Are you fine with any of the options above? Do you have a better idea?<br>
>>>><br>
>>>> -Eli<br>
>>><br>
>><br>
><br>
><br>
</div></div></blockquote></div><br>