libc/stdio: Increase BUF in vfprintf.c and vfwprintf.c

[VexedUXR]

With the new %b format specifier, we need at most 64 characters.

Also, the comment says that we need enough space to write a uintmax_t in octal plus one byte, I'm not sure what that's for. I guess its implying that we need a null terminator(?), but it doesn't look like we do.

This can be easily tested by doing:

printf("%lb\n", 1ul << n); /* Where n is anywhere from 32 to 63 */

[VexedUXR]

Thinking about this some more, we should just be able to remove the whole conditional and make BUF sizeof(uintmax_t) * CHAR_BIT

[bsdimp]

Don't you need +2 for the 0b?

[VexedUXR]

Oh, I didn't think of that, but it looks like we print it separately:

freebsd-src/lib/libc/stdio/vfprintf.c

Line 1022 in 6e41473

if (ox[1]) { /* ox[1] is either x, X, or \0 */

freebsd-src/lib/libc/stdio/vfwprintf.c

Line 1077 in 6e41473

if (ox[1]) { /* ox[1] is either x, X, or \0 */

[bsdimp]

Then don't we need +1 for NUL?

[VexedUXR]

As far as I can tell, we don't.

In the case of %c, we write to it manually (or in the case of a wide char through wcrtomb). Otherwise we pass it to __ujtoa or __ultoa which both stop when they're done writing the number.
After that we just print with size which is calculated using buf + BUF - cp where cp is buf plus some offset.

When calling grouping_print, to print a number with separators, we pass a start pointer and end pointer (cp and buf+BUF), which are used to call io_print (through io_printandpad) with the correct size.

Overall the code looks like it doesn't expect a null terminator. However, I think that the comment I removed was hinting that we did, otherwise I'm not sure what that +1 would be for. Unless there's some edge case I'm missing...

FWIW, the patch also survived the cirrus-ci boot smoke test.

[bsdimp]

I think that the 0b that's prented isn't accounted for, but isn't now.