[PATCH] Fix UBSAN in __ubsan::Value::getSIntValue

[marxin]

Summary

As reported here:
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=99190

it fixes:

// gcc -fsanitize=shift -w
// ldd a.out produces
// libubsan.so.1 => /home/vitti/local/gcc-150221-undefined/lib64/libubsan.so.1
// ./a.out produces
// ../../../../gcc-150221/libsanitizer/ubsan/ubsan_value.cpp:77:25: runtime error: left shift of 0x0000000000000000fffffffffffffffb by 96 places cannot be represented in type '__int128'
// ubsan_value.cpp:77 -> "return SIntMax(Val) << ExtraBits >> ExtraBits;"
int main (void)
{

  1 << -5;

  return 0;
}

[vitalybuka]

Can you try to add a test?

[marxin]

In D97263#2581226, @vitalybuka wrote:

Can you try to add a test?

I don't an easy way how to do that. Anyway, I've just pushed the revision.
Thanks for the review.

compiler-rt/lib/ubsan/ubsan_value.cpp

 
 SIntMax Value::getSIntValue() const {
   CHECK(getType().isSignedIntegerTy());
   if (isInlineInt()) {
     // Val was zero-extended to ValueHandle. Sign-extend from original width
     // to SIntMax.
     const unsigned ExtraBits =
       sizeof(SIntMax) * 8 - getType().getIntegerBitWidth();
-    return SIntMax(Val) << ExtraBits >> ExtraBits;
+    return SIntMax(UIntMax(Val) << ExtraBits) >> ExtraBits;
   }
   if (getType().getIntegerBitWidth() == 64)
     return *reinterpret_cast<s64*>(Val);
 #if HAVE_INT128_T
   if (getType().getIntegerBitWidth() == 128)
     return *reinterpret_cast<s128*>(Val);
 #else
   if (getType().getIntegerBitWidth() == 128)