64-bit word literals are parsed incorrectly on 32-bit systems

[turboencabulator]

Version

110.99.3 (Latest)

Operating System

• Any
• Linux
• macOS
• Windows
• Other Unix

OS Version

No response

Processor

• Any
• Arm (using Rosetta)
• PowerPC
• Sparc
• x86 (32-bit)
• x86-64 (64-bit)
• Other

System Component

Basis Library

Severity

Major

Description

Bits 30 and 31 of 64-bit word literals are lost on a 32-bit machine. The problem occurs with version 110.98.1 and later.

Transcript

Standard ML of New Jersey (32-bit) v110.99.3 [built: Tue Oct 18 00:40:03 2022]
- 0wxFFFFFFFFFFFFFFFF : Word64.word;
[autoloading]
[library $SMLNJ-BASIS/basis.cm is stable]
[library $SMLNJ-BASIS/(basis.cm):basis-common.cm is stable]
[autoloading done]
val it = 0wxFFFFFFFF3FFFFFFF : Word64.word

Expected Behavior

No response

Steps to Reproduce

0wxFFFFFFFFFFFFFFFF : Word64.word

Additional Information

No response

Email address

elyk03@gmail.com

[JohnReppy]

Note that this is an issue with the 32-bit version of SML/NJ; it is not present in the 64-bit version.

[turboencabulator]

I found another issue with string matching, also only affects 32-bit, and also started in 110.98.1:

fun f "abcdefg" = true
  | f _ = false;
f "abcdefg";  (* ok *)
f "abcdef_";  (* bad *)
f "abcde__";  (* bad *)
f "abcd___";  (* bad *)
f "abc____";  (* ok *)

Looking into it further, I think I've found the root cause of both this and the original issue.
base/compiler/CPS/opt/streqlcnv.sml:

    val wordSzInBytes = if Target.is64 then 8 else 4
    val bytesToWords = if Target.is64
	  then (fn n => Word.toIntX(Word.>>(Word.fromInt n + 0w7, 0w3)))
	  else (fn n => Word.toIntX(Word.>>(Word.fromInt n + 0w7, 0w3)))

The last line wasn't updated for 32-bit.

[JohnReppy]

Thanks for the pointing out the source of the problem. I have committed a fix for 110.99.4.