[PATCH]: Implement fwide (was Re: swprintf() and friends?)
Jeff Johnston
jjohnstn@redhat.com
Thu Dec 11 20:11:00 GMT 2008
Just to confirm that modified patch checked in.
-- Jeff J.
Jeff Johnston wrote:
> Corinna Vinschen wrote:
>> On Dec 10 15:40, Jeff Johnston wrote:
>>
>>> Corinna Vinschen wrote:
>>>
>>>> Jeff?
>>>>
>>>> Did you have a look in the meantime?
>>>>
>>>>
>>> I actually was looking at this today.
>>>
>>> Couple of immediate points:
>>>
>>> 1. In ORIENT, you do a fp->flags != __SORD (should be "|= ")
>>>
>>
>> Ouch! Thanks for spotting.
>>
>>
>>> 2. ORIENT should not check the value of ori. It is only 0 in one
>>> case when fwide
>>> is used so fwide should check for 0 and not call ORIENT in that
>>> case.
>>>
>>
>> Fine with me.
>>
>>
>>> 3. There doesn't seem to be orientation for the printf family while
>>> the scanf family is covered
>>> via vfscanf.
>>>
>>
>> Hmm, you seem to be right. I examined the FreeBSD sources again since I
>> was pretty sure that the code path will result in calling ORIENT when
>> calling any printf function. But it really looks like printf doesn't
>> set the orientation. A bug in the FreeBSD sources? __sfvwrite_r might
>> be a good point to call ORIENT, right?
>>
>>
> I decided just to put it in _VFPRINT_R after the lock. Seemed
> consistent.
>>> I can fix these myself if you like or you can resubmit if you would
>>> like to test.
>>>
>>
>> I have no problems if you fix that.
>>
>>
> Done. I also changed references to stdin and stdout in wchar.h so
> they basically expand as they do in stdio.h.
>
> I think Craig's idea to move the functions to a wchar directory is a
> reasonable idea. I am going to check the code in the stdio directory
> for now as we are really close to the snapshot and I have other things
> to do. I want people to have a chance to find any problems before the
> snapshot and if I don't get a chance to do the move or run into any
> problems, the functions will be available.
>
> Thanks Corinna for doing this.
>> Thanks,
>> Corinna
>>
>>
>
More information about the Newlib
mailing list