[PATCH] D157079: [TableGen] Fix wrong lex result on 64-bit integer boundaries

[Fangrui Song via Phabricator via llvm-commits]

MaskRay accepted this revision.
MaskRay added a comment.
This revision is now accepted and ready to land.

LGTM.



================
Comment at: llvm/lib/TableGen/TGLexer.cpp:468
+
+  // Check if it's a hex or a binary value
   if (CurPtr[-1] == '0') {
----------------



================
Comment at: llvm/lib/TableGen/TGLexer.cpp:484
 
-  // Check for a sign without a digit.
-  if (!isdigit(CurPtr[0])) {
-    if (CurPtr[-1] == '-')
-      return tgtok::minus;
-    else if (CurPtr[-1] == '+')
-      return tgtok::plus;
+  // For a hex or binary value, we always convert it to an unsigned value
+  bool IsMinus = false;
----------------
Add a period to a full sentence.


================
Comment at: llvm/test/TableGen/64-bit-int.td:22
+#ifdef VALID
+  bits<64> BinVal = 0x8000000000000000;
+  bits<64> HexVal = 0b1000000000000000000000000000000000000000000000000000000000000000;
----------------
I think you swapped BinVal and HexVal?

I think it's useful to give the same suffix for Bin/Hex/Dec values with the same value.

```
BinVal0 = 0b.........;
HexVal0 = 0x8000000000000000;
bits<64> DecVal0 = 9223372036854775808;

HexVal1 = 0x8000000000000001;
bits<64> DecVal1 = 9223372036854775809;
```

This grouping makes it clear that  9223372036854775808 equals 0x8000000000000000....


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D157079/new/

https://reviews.llvm.org/D157079