Compile performance of Linux kernels in mainline gcc

[Daniel Berlin]

On Oct 30, 2004, at 10:11 AM, Andi Kleen wrote:

> On Sat, Oct 30, 2004 at 10:09:29AM -0400, Andrew Pinski wrote:
>>
>> On Oct 30, 2004, at 10:07 AM, Andi Kleen wrote:
>>
>>> On Sat, Oct 30, 2004 at 09:48:53AM -0400, Andrew Pinski wrote:
>>>>
>>>> Yes the problem is that his terminal does not support UTF8 and his
>>>> LC_* are set to UTF8, a non bug.
>>>
>>> Huh?  While that's possible why does gcc need UTF-8 to print 
>>> standard C
>>> identifiers?
>>
>> It is printing quotes, not the identifier itself.
>
> Quotes? You mean >>"<< ?  That is 7 bit ASCII too.

On most planets, yes.
However, in UTF-8 world, we apparently print nice quotes, not >>>"<<<<.
(I'd put them here, but they would obviously display as garbage to you, 
as we've determined ;P)

I discovered this running the testsuite once without LANG=C, and seeing 
tons of failures because the quotes look different than the strings we 
were expecting.
 
>
>
> -Andi

On Oct 30, 2004, at 10:11 AM, Andi Kleen wrote:

On Sat, Oct 30, 2004 at 10:09:29AM -0400, Andrew Pinski wrote:

On Oct 30, 2004, at 10:07 AM, Andi Kleen wrote:

On Sat, Oct 30, 2004 at 09:48:53AM -0400, Andrew Pinski wrote:

Yes the problem is that his terminal does not support UTF8 and his
LC_* are set to UTF8, a non bug.

Huh?  While that's possible why does gcc need UTF-8 to print standard C
identifiers?

It is printing quotes, not the identifier itself.

Quotes? You mean >>"<< ?  That is 7 bit ASCII too.

On most planets, yes.
However, in UTF-8 world, we apparently print nice quotes, not >>>"<<<<.
(I'd put them here, but they would obviously display as garbage to you, 
as we've determined ;P)

I discovered this running the testsuite once without LANG=C, and seeing 
tons of failures because the quotes look different than the strings we 
were expecting.
 


-Andi