Fix yet one data race in PCRE

PCRE 8.x initializes the pattern compiler on demand during the first
pcre_study call. It could be worse, but since the compiled patterns are
cached, the locking impact is minimal. PCRE 10.x always compiles the
pattern and thread sanitizer doesn't complain about the compiler
initialization, thus the newer PCRE version seems to be unafected.

ext/pcre/php_pcre.c

@@ -67,7 +67,7 @@ PHPAPI ZEND_DECLARE_MODULE_GLOBALS(pcre)
 #define PCRE_JIT_STACK_MAX_SIZE (64 * 1024)
 ZEND_TLS pcre_jit_stack *jit_stack = NULL;
 #endif
-#if defined(ZTS) && defined(HAVE_PCRE_JIT_SUPPORT)
+#if defined(ZTS)
 static MUTEX_T pcre_mt = NULL;
 #define php_pcre_mutex_alloc() if (!pcre_mt) pcre_mt = tsrm_mutex_alloc();
 #define php_pcre_mutex_free() if (pcre_mt) tsrm_mutex_free(pcre_mt); pcre_mt = NULL;
@@ -538,7 +538,9 @@ PHPAPI pcre_cache_entry* pcre_get_compiled_regex_cache(zend_string *regex)
 	/* If study option was specified, study the pattern and
 	   store the result in extra for passing to pcre_exec. */
 	if (do_study) {
+		php_pcre_mutex_lock();
 		extra = pcre_study(re, soptions, &error);
+		php_pcre_mutex_unlock();
 		if (extra) {
 			extra->flags |= PCRE_EXTRA_MATCH_LIMIT | PCRE_EXTRA_MATCH_LIMIT_RECURSION;
 			extra->match_limit = (unsigned long)PCRE_G(backtrack_limit);