Bug#28008107 MAKE UBSAN ERROR REPORTING FAIL-FAST

Post-push fix. Some callers of dynstr_append_mem() apparently depend
on the side-effects of zero-termination when appending a zero-sized string.
This is from clang version 6.0.0-1ubuntu2 with ASAN:

==25852==ERROR: AddressSanitizer: heap-buffer-overflow on address 0x604000189d40 at pc 0x0000004e076c bp 0x7ffd77220550 sp 0x7ffd7721fd00
READ of size 17 at 0x604000189d40 thread T0
    #0 0x4e076b in __interceptor_strlen.part.30 asan_interceptors.cc.o:?
    mysql#1 0x5b0496 in init_dynamic_string(DYNAMIC_STRING*, char const*, unsigned long, unsigned long) mysys/my_string.cc:51
    mysql#2 0x56accd in do_connect(st_command*) client/mysqltest.cc:5970
    mysql#3 0x5680c9 in main client/mysqltest.cc:8945
    mysql#4 0x7f027e87db96 in __libc_start_main /build/glibc-OTsEL5/glibc-2.27/csu/../csu/libc-start.c:310

The fix is to skip memcpy() only, if length == zero, rather than
skipping the entire function.

Change-Id: I10540d7c87ae1cfed3bc6967e42290401f5d1f8d

Author: Tor Didriksen

mysys/my_string.cc

@@ -106,7 +106,6 @@ bool dynstr_append(DYNAMIC_STRING *str, const char *append) {
 
 bool dynstr_append_mem(DYNAMIC_STRING *str, const char *append, size_t length) {
   char *new_ptr;
-  if (length == 0) return false;
   if (str->length + length >= str->max_length) {
     size_t new_length =
         (str->length + length + str->alloc_increment) / str->alloc_increment;
@@ -117,7 +116,7 @@ bool dynstr_append_mem(DYNAMIC_STRING *str, const char *append, size_t length) {
     str->str = new_ptr;
     str->max_length = new_length;
   }
-  memcpy(str->str + str->length, append, length);
+  if (length > 0) memcpy(str->str + str->length, append, length);
   str->length += length;
   str->str[str->length] = 0; /* Safety for C programs */
   return false;