[compiler-rt][MSAN][test] Update buffersize for UTF-8 to C convert

[jsji]

Summary

This is exposed by https://reviews.llvm.org/D83486.
When the host is UTF8, we may get n >10, causing assert failure.
Increase the buffersize to support UTF-8 to C conversion.

[broadwaylamb]

Could you please explain why D83486 broke this test and what does PowerPC have to do with it? I’m just curious.

Ah, and thanks for fixing the issue!

[jsji]

In D83719#2148520, @broadwaylamb wrote:

Could you please explain why D83486 broke this test and what does PowerPC have to do with it? I’m just curious.

Ah, and thanks for fixing the issue!

Actually, it has nothing to do with PowerPC, just happens on our bots.

D83486 change the lit test env, to use parent process env, which include locale setting, and our PowerPC bots have UTF-8 locale by default,
so when we call newlocale in this test, we get UTF-8 locale instead of C locale.

So when we call strxfrm_l we get n =20 > 10, so the tests assert in line 19..

compiler-rt/test/msan/strxfrm.cpp

 // RUN: %clangxx_msan -O0 -g %s -o %t && %run %t
 
 #include <assert.h>
 #include <locale.h>
 #include <sanitizer/msan_interface.h>
 #include <stdlib.h>
 #include <string.h>
 
 int main(void) {
-  char q[10];
+  char q[30];
   size_t n = strxfrm(q, "abcdef", sizeof(q));
   assert(n < sizeof(q));
   __msan_check_mem_is_initialized(q, n + 1);
 
   locale_t loc = newlocale(LC_ALL_MASK, "", (locale_t)0);
 
   __msan_poison(&q, sizeof(q));
   n = strxfrm_l(q, "qwerty", sizeof(q), loc);
   assert(n < sizeof(q));
   __msan_check_mem_is_initialized(q, n + 1);
   return 0;
 }