homepage

Steps 1 and 3 of PyUnicode_FromFormatV() doesn't handle the format string "%%" correctly. The loop responsible to skip the precision moves outside the format string, and the function will then read uninitialized memory. The loop:
 while (*++f && *f != '%' && !Py_ISALPHA((unsigned)*f))
 ;
This is another issue:
 for (f = format; *f; f++) {
 if (*f == '%') {
 if (*(f+1)=='%')
 continue;
 ...
continue only skips the first %: with "%%", the second % will be interpreted (and not escaped).
Attached patch fixes the issue, but I don't feal confortable with this ugly function, and I would appreciate a review :-) The patch adds unit tests.
I found the bug when trying to add new tests before trying to implement "%zi" format. I was first surprised that "%zi" (and %li and %lli) is not supported, but now I am surprised because I found bugs :-)