[patch] Fix BZ 19165 -- overflow in fread / fwrite

[Paul Eggert]

On 02/11/2016 07:58 AM, Zack Weinberg wrote:
> I don't think anyone wants the behavior to vary depending on which
> kind of fd you have passed to read().

As an application developer I'd rather have 'read', 'fread', etc. fail, 
or even dump core, if I give them a buffer that is not entirely valid. I 
hope, for example, that checkers like valgrind object to this sort of 
thing, and would find it useful if glibc continues to report it as an 
error. If POSIX really requires that 'read' and 'fread' must succeed in 
this situation, then from my point of view it's a POSIX bug that needs 
to get fixed, as the benefit of the extra checking is significant.