Add global stdio streams support for reent small.

[Douglas]

Hi Jeff,

Updated the comments. The logic for those does not appear to be
consistent in the source code - should it use negative logic on the
#else or repeat the #if logic?  An attempt was made to follow the local
conventions in each file, and if you have a preference for negating the
logic on the #else then just let me know?

Re: findfp.c and the lock initialization. The original logic was:

!defined(__SINGLE_THREAD__) && !defined(_REENT_SMALL)

so the lock was not initialized if:

__SINGLE_THREAD__ || _REENT_SMALL

Then using _REENT_SMALL and the _REENT_GLOBAL_STDIO_STREAMS this lock
needs to be initialized here, so an exception for that was added:

!defined(__SINGLE_THREAD__) && !(defined(_REENT_SMALL) &&
!defined(_REENT_GLOBAL_STDIO_STREAMS))

and the lock is not initialized if:

__SINGLE_THREAD__ || (_REENT_SMALL && !_REENT_GLOBAL_STDIO_STREAMS))

Currently _REENT_GLOBAL_STDIO_STREAMS does not work with _REENT_SMALL so
this change does not appear to break that case - this patch is to fix
the case in which these are both used together.

If I have missed a logic case there could you please point to the
combination of options that it does break so I can see that clearly, thanks.

Douglas

On 03/02/2018 07:22 AM, Jeff Johnston wrote:
> Hi Douglas,
> 
> A few comments.  First of all, please update comments on preprocessor #if,
> #else statements to
> include your changes (e.g. you have added a #else, what does this #else
> mean...e.g. #else /* _REENT_SMALL || _REENT_GLOBAL_STDIO_STREAMS */
> This will aid future maintainers to not have to look all the way back to
> the original #if).  This also applies to
> comments for #else statements that are altered by your changes.
> 
> That said, I have one code comment.
> 
> In findfp.c std, you added "&& !defined(__REENT_GLOBAL_STDIO_STREAMS)" to
> check if the lock for the file should be initialized.  I don't understand
> why you added this.  AFAICT, this will affect current behaviour.
> 
> Regarding slots in the reent structs, yes, they are sometimes there for
> fill, for backwards compatibility.  Otherwise, everything needs to be
> recompiled/relinked.
> 
> -- Jeff J.
> 
> On Wed, Feb 28, 2018 at 7:58 AM, Douglas <info@ourairquality.org> wrote:
> 
>> The global stdio streams option did not appear to work with reent small
>> option, but these two options are both useful on a low memory system.
>>
>> This seems to simplify the reent small code in this area when using the
>> global stdio streams option. The fake streams and some of the checking
>> is not longer necessary.
>>
>> btw the __sf slot in the reent small struct _reent appears to be unused.
>> Could that be removed, or is it fill for backwards compatibility?
>>
> 

-------------- next part --------------
A non-text attachment was scrubbed...
Name: 0001-Add-global-stdio-streams-support-for-reent-small.patch
Type: text/x-patch
Size: 5474 bytes
Desc: not available
URL: <http://sourceware.org/pipermail/newlib/attachments/20180301/3a41343a/attachment.bin>