[sanitizer] Always initialize the regex in the regcomp() interceptor

[tavianator]

Summary

When regcomp() fails, the same regex_t* should be passed to regerror()
for potentially better error messages. But doing that with msan would
report a use-of-uninitialized-value.

Fixes https://github.com/google/sanitizers/issues/1496

[vitalybuka]

Thank you!
I can land it later, if you have no access.

compiler-rt/lib/sanitizer_common/sanitizer_common_interceptors.inc

7978

compiler-rt/test/sanitizer_common/TestCases/Posix/regex.cpp

44–48

you don't need to move them, just use {} to make sure it's two different instances

[tavianator]

Address review comments

[vitalybuka]

compiler-rt/lib/sanitizer_common/sanitizer_common_interceptors.inc

7977–7978

FYI: the patch is upload somehow incorrectly, "before" does not match existing code. if (!res) is missing.
Arcanist, as explained here https://llvm.org/docs/Phabricator.html#id7, is easy way to upload it correctly.
No need to fix this one, I resolved conflicts already.

[vitalybuka]

fixed conflicts

compiler-rt/lib/sanitizer_common/sanitizer_common_interceptors.inc

 
 #if SANITIZER_INTERCEPT_REGEX
 INTERCEPTOR(int, regcomp, void *preg, const char *pattern, int cflags) {
   void *ctx;
   COMMON_INTERCEPTOR_ENTER(ctx, regcomp, preg, pattern, cflags);
   if (pattern)
     COMMON_INTERCEPTOR_READ_RANGE(ctx, pattern, internal_strlen(pattern) + 1);
   int res = REAL(regcomp)(preg, pattern, cflags);
-  if (!res)
+  if (preg)
     COMMON_INTERCEPTOR_WRITE_RANGE(ctx, preg, struct_regex_sz);
   return res;
 }
 INTERCEPTOR(int, regexec, const void *preg, const char *string, SIZE_T nmatch,
             struct __sanitizer_regmatch *pmatch[], int eflags) {
   void *ctx;
   COMMON_INTERCEPTOR_ENTER(ctx, regexec, preg, string, nmatch, pmatch, eflags);
   if (preg)
     COMMON_INTERCEPTOR_READ_RANGE(ctx, preg, struct_regex_sz);

compiler-rt/test/sanitizer_common/TestCases/Posix/regex.cpp

void test_print_matches(const regex_t *preg, const char *string) {
   } else if (rv == REG_NOMATCH)
     printf("%s: not-matched\n", string);
   else
     abort();
 }
 
 int main(void) {
   printf("regex\n");
 
+  {
     regex_t regex;
     int rv = regcomp(&regex, "[[:upper:]]\\([[:upper:]]\\)", 0);
     assert(!rv);
 
     test_matched(&regex, "abc");
     test_matched(&regex, "ABC");
 
     test_print_matches(&regex, "ABC");
 
     regfree(&regex);
+  }
 
-  rv = regcomp(&regex, "[[:upp:]]", 0);
+  {
+    regex_t regex;
+    int rv = regcomp(&regex, "[[:upp:]]", 0);
     assert(rv);
 
     char errbuf[1024];
     regerror(rv, &regex, errbuf, sizeof errbuf);
     printf("error: %s\n", errbuf);
 
+    regfree(&regex);
+  }
+
   // CHECK: regex
   // CHECK: abc: not-matched
   // CHECK: ABC: matched
   // CHECK: matched[0]='AB'
   // CHECK: matched[1]='B'
   // CHECK: error:{{.*}}
 
   return 0;
 }