Handle trailing % in strftime (#23045)

"In glibc a trailing % in strftime() acts like printf, ie it's a literal
%". This is breaking some Python tests. I've applied the patch suggested
here: https://www.openwall.com/lists/musl/2022/12/19/2

Author: hoodmane

system/lib/libc/README.md

@@ -12,6 +12,7 @@ Some changes have been made to the version that was taken from upstream, includi
  * Switch to using the wasi `fd_write` syscall instead of `writev`.
  * Simplify stdout stream handling: do not support seeking, terminal handling, etc., as it just increases code size and Emscripten doesn't have those features anyhow.
  * Setting `_POSIX_REALTIME_SIGNALS` and `_POSIX_SPAWN` macros to -1, to exclude unsupported functions.
+ * Handling trailing % in `strftime` and `wcsftime` format strings.
 
 Copy log.c and log2.c from earlier version of musl which result in smaller
 binary size since they do not rely data tables in log_data.c and log2_data.c.

system/lib/libc/musl/src/time/strftime.c

@@ -226,7 +226,13 @@ size_t __strftime_l(char *restrict s, size_t n, const char *restrict f, const st
 			s[l] = 0;
 			return l;
 		}
+#ifdef __EMSCRIPTEN__
+		// Handle trailing % by outputting a % rather than returning 0. Ideally
+		// this 6 character change could be upstreamed into musl...
+		if (*f != '%' || !f[1]) {
+#else
 		if (*f != '%') {
+#endif
 			s[l++] = *f;
 			continue;
 		}

test/other/test_strftime.c

@@ -307,5 +307,8 @@ int main() {
   size = strftime(s, sizeof(s), "%Ec", &tm);
   TEST(!cmp(s, "Mon Dec 17 00:00:00 2018"), "strftime test #36a", s);
 
+  size = strftime(s, sizeof(s), "trailing %", &tm);
+  TEST((size == 10), "strftime test #37", s);
+  TEST(!cmp(s, "trailing %"), "strftime test #37", s);
   return 0;
 }