[prev in list] [next in list] [prev in thread] [next in thread] 

List:       kde-devel
Subject:    Re: Byte order conversions
From:       Szombathelyi_György <gyurco () freemail ! hu>
Date:       2004-10-28 7:00:31
Message-ID: 4180990F.50108 () freemail ! hu
[Download RAW message or body]

Hello!

Richard Smith wrote:
> 
> My 2¢:
> 
> The KBE2ME_nn macros and the like could just as well be inline functions. 
> That'd be better for debugging; you'd be able to call them from gdb, and step 
> into them. I personally would prefer more readable/obvious function/macro 
> names.
> 
I hope these functions doesn't need any debugging. In contrast I don't 
want names like KBigEndianToMachineEndian. Personal taste, I think. 
Actually I took the function names from MPlayer's own bswap.h and 
prefixed them with 'K'.

> I think the KSWAP_nn functions would be more readable if they were made less 
> compact; there's nothing wrong with having a brace on a line of its own 
> (personal preference of course, YMMV). 
I'm ok with this.

Again (but less important this time)
> I'd prefer more verbose names. And making them static will result in huge 
> bloat if the compiler decides to make out-of-line copies (eg for debugging 
> purposes).
> 
These functions are so small, is that really matters?

> And a templated machineToBigEndian<T>(...), bigToMachineEndian<T>(...), etc 
> would be nice too (so the caller doesn't have to care how wide an int is, for 
> instance).

The places where these functions are used are very sensitive to the 
length of the actual data, so one should always know its size, I think.

Thanks for the comments!

Bye,
György
 
>> Visit http://mail.kde.org/mailman/listinfo/kde-devel#unsub to unsubscribe <<
[prev in list] [next in list] [prev in thread] [next in thread]