On 11/17/2010 11:39 PM, Chris Adams wrote:
> Once upon a time, Matthew Garrett <mjg59@xxxxxxxxxxxxx> said:
>> On Wed, Nov 17, 2010 at 10:30:00PM -0600, Chris Adams wrote:
>>> How is that relevant? If the behavior changes on only some
>>> architectures, then it is okay?
>>
>> If it's broken on non-x86 already then there haven't been "years of 'it
>> just works'".
>
> It doesn't change the fact that glibc changed behavior on some CPUs
> (which is also going to make tracking down memcpy bugs highly
> irritating, since the behavior will be different depending on the CPU).
>
> Any normal person writing code is going to write a memcpy that copies
> up, whether a simple C loop or optimized assembly, so I really doubt
> you'll find lots of architectures that are widely used in the Unix world
> where people use the memcpy routine in the standard C library that does
> something different.
If you code it to copy up instead of down, then it just means you've opted
to misbehave in the /other/ overlap case instead of this one. There's no
winning to be had here.
--
Peter
When in doubt, debug-on-entry the function you least suspect has
anything to do with something.
--
devel mailing list
devel@xxxxxxxxxxxxxxxxxxxxxxx
https://admin.fedoraproject.org/mailman/listinfo/devel
