flac dev - Oct 2015 - [PATCH] cpu.c: initialize a variable

Some debugging tools don't like if a code uses unitialized variables.
The code in cpu.c uses the value of fxsr.buff[50], so it makes sense
to explicitely initialize it (because fxsave can fail to do it).
-------------- next part --------------
A non-text attachment was scrubbed...
Name: cpu_init_val.patch
Type: application/octet-stream
Size: 409 bytes
Desc: not available
Url :
http://lists.xiph.org/pipermail/flac-dev/attachments/20151012/2f4eb2e0/attachment.obj

lvqcl wrote:
> Some debugging tools don't like if a code uses unitialized variables.
> The code in cpu.c uses the value of fxsr.buff[50], so it makes sense
> to explicitely initialize it (because fxsave can fail to do it).
Your change:

		fxsr.buff[50] = 0;

only changes one element in that 128 work buffer. Why explicitly initialize
onky the 50th element and not all of them?

My approach would be:

     memset (fxsr.buff, 0, sizeof (fxsr.buff));

What do you think?

Erik
-- 
----------------------------------------------------------------------
Erik de Castro Lopo
http://www.mega-nerd.com/

Erik de Castro Lopo wrote:
>> Some debugging tools don't like if a code uses unitialized
variables.
>> The code in cpu.c uses the value of fxsr.buff[50], so it makes sense
>> to explicitely initialize it (because fxsave can fail to do it).
>
> Your change:
>
> 		fxsr.buff[50] = 0;
>
> only changes one element in that 128 work buffer. Why explicitly initialize
> onky the 50th element and not all of them?
The C code that follows:

     ...
     old_val = fxsr.buff[50];
     fxsr.buff[50] ^= 0x0013c0de;
     ...
     fxsr.buff[50] = old_val;
     ...
     new_val = fxsr.buff[50];
     fxsr.buff[50] = old_val;
     ...

uses only fxsr.buff[50], so IMHO it's enough to initialize only the value
of this element.

> My approach would be:
>
>      memset (fxsr.buff, 0, sizeof (fxsr.buff));
>
> What do you think?
It won't hurt but probably superfluous.