[Bug-gnulib] Re: snprintf

[Simon Josefsson]

Bruno Haible <address@hidden> writes:

> Simon Josefsson wrote:
>> I don't understand, could you be more specific?  snprintf still need
>> to return the full length, so it has to be computed somehow.
>
> What I meant is this. Does this look OK?

I not sure.

If vasnprintf really doesn't realloc the input string (which I also
thought initially), that should be mentioned in the vasnprintf
documentation (it doesn't give that impression now).  It seems like a
critical interface dependency.  Do all vasnprintf implementations
behave the same?  If not, do the vasnprintf M4 macro test for it?

Also, if size==0, it seems your version might dereference STR (inside
vasnprintf), which also seems bad.  I'm assuming snprintf(4711, 0,
"foo") should not access memory at 4711, but still return 3.

Thanks.