Jim Meyering <[EMAIL PROTECTED]> writes: > vasnprintf.c should resort to incurring the > expense of that portability kludge only when necessary. > How about if it does that only when a configure-time test > has detected that snprintf's return value is not conforming?
That sounds like a good idea. How common is it for snprintf to return the wrong value? If it's not that common, perhaps we can just not worry about those systems at all. I can't resist mentioning a more-fun workaround. We can put a snprintf wrapper into gnulib that copies the format string into a buffer, uses mmap to make the buffer read-only, and then calls the real snprintf. Sort of an "anti-Fortify" wrapper, if you will.
