Re: [PATCH] plugin_ops.h fixes

[Takashi Iwai <tiwai@suse.de>]

At Mon, 08 Jul 2002 11:02:25 +0200,
Clemens Ladisch wrote:
> 
> This macro in plugin_ops.h
> 	#ifdef __i386__
> 	#define _get_triple_le(ptr) (*(u_int32_t*)(ptr) & 0xffffff)
> tries to access four bytes through the pointer. If the three bytes of
> valid data are at the end of a page and the next page isn't mapped,
> this results in an exception.

right...

> The companion macro
> 	#define _get_triple_be(ptr) (bswap_32(*(u_int32_t*)(ptr)) & 0xffffff)
> suffers from the same problem, and, additionally, is wrong (it lacks
> ">> 8" after the bswap).
> I fixed both macros by removing them. :-)
 
yes, thanks.


> There is a typo in get16_1230_B2.
> 
> The put16_labels array is declared as having 128 elements instead
> of 16.
> 
> The gets_* code doesn't correctly extend the sign of the value in every
> case.
> 
> put_12_A1 should have been put_12_29, and the codes for put_0123_* use
> the nonexistent macros as_s24/as_u24. (This doesn't really matter as
> put_* isn't used anyway.)
> 
> The _norms function appears to be badly broken (_min and _max
> interchanged, using 32/64 bits for 24/32 bit values). I didn't attempt
> to fix it as it isn't used, either.
 
yes.  we can remove this stuff.
Jaroslav, any plan to use still this one?

> And the summing/normalization code in pcm_route.c treats the sample as
> if it were unsigned.

nice spot.

patches are applied to cvs.


thanks,

Takashi


-------------------------------------------------------
This sf.net email is sponsored by:ThinkGeek
Oh, it's good to be a geek.
http://thinkgeek.com/sf