Re: [PATCH 3/3] sha1: use char type for temporary work buffer

[Yann Droneaud <ydroneaud@opteya.com>]

Le mercredi 12 septembre 2012 à 14:38 -0400, Jeff King a écrit :
> On Wed, Sep 12, 2012 at 12:30:45PM +0200, Yann Droneaud wrote:
> 
> > The SHA context is holding a temporary buffer for partial block.
> > 
> > This block must 64 bytes long. It is currently described as
> > an array of 16 integers.
> > 
> > Signed-off-by: Yann Droneaud <ydroneaud@opteya.com>
> > ---
> >  block-sha1/sha1.h | 2 +-
> >  1 file changed, 1 insertion(+), 1 deletion(-)
> > 
> > diff --git a/block-sha1/sha1.h b/block-sha1/sha1.h
> > index b864df6..d29ff6a 100644
> > --- a/block-sha1/sha1.h
> > +++ b/block-sha1/sha1.h
> > @@ -9,7 +9,7 @@
> >  typedef struct {
> >  	unsigned long long size;
> >  	unsigned int H[5];
> > -	unsigned int W[16];
> > +	unsigned char W[64];
> >  } blk_SHA_CTX;
> 
> Wouldn't this break all of the code that is planning to index "W" by
> 32-bit words (see the definitions of setW in block-sha1/sha1.c)?
> 

That's not the same "W" ... This part of the code is indeed unclear.

> You do not describe an actual problem in the commit message, but reading
> between the lines it would be "system X would like to use block-sha1,
> but has an "unsigned int" that is not 32 bits". IOW, an ILP64 type of
> architecture. Do you have some specific platform in mind?
> 
> If that is indeed the problem, wouldn't the simplest fix be using
> uint32_t instead of "unsigned int"?
> 

It's another way to fix this oddity, but not simpler.


> Moreover, would that be sufficient to run on such a platform? At the
> very least, "H" above would want the same treatment. And I would not be
> surprised if some of the actual code in block-sha1/sha1.c needed
> updating, as well.
> 

ctx->H is actually used as an array of integer, so it would benefits of
being declared uint32_t for an ILP64 system. This fix would also be
required for blk_SHA1_Block() function.

Regards.

-- 
Yann Droneaud
OPTEYA