Fix sign-extension of 64-bit literals#213
Open
laurenthuberdeau wants to merge 2 commits into
Open
Conversation
When SUPPORT_64_BIT_LITERALS is not enabled, INTEGER nodes store the negation of the integer value. This is done to support the full range of 32-bit signed integers, since the positive value 2147483648 cannot be represented as a positive signed 32-bit integer. To get the actual value, it must be negated. If the negated value is -2147483648, negating it produces 2147483648, which overflows the 32-bit signed integer range (assuming the host platform uses 32-bit integers) and loops back to -2147483648. Additionally, for 64-bit targets, the value must be sign-extended to 64-bits, which is done by left shifting by 31. On platforms with 32-bit integers, the two's complement representation of 2147483648 is the same as -2147483648's two's complement, 0x80000000 in hex, and so everything works out. On platforms with larger integers, such as some shells, the value 2147483648 is representable as a positive integer, and so negating -2147483648 produces 2147483648, which produces an invalid sign-extended value when shifted. Sign-extension is only performed when targeting 64-bit platforms, where 2147483648 >> 31 evaluates to 1.
laurenthuberdeau
added a commit
that referenced
this pull request
Nov 11, 2025
Without this, the 0x80000000 literal in sign-extended which causes differences in the bootstrapped compiler binaries. See #213 for more information.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Context
When
SUPPORT_64_BIT_LITERALSis not enabled,INTEGERnodes store the negation of the integer value. This is done to support the full range of 32-bit signed integers, since the positive value2147483648cannot be represented as a positive signed 32-bit integer.To get the actual value, it must be negated. If the negated value is
-2147483648, negating it produces2147483648, which overflows the 32-bit signed integer range (assuming the host platform uses 32-bit integers) and loops back to-2147483648. Additionally, for 64-bit targets, the value must be sign-extended to 64-bits, which is done by left shifting by 31.On platforms with 32-bit integers, the two's complement representation of 2147483648 is the same as -2147483648's two's complement, 0x80000000 in hex, and so everything works out.
On platforms with larger integers, such as some shells, the value
2147483648is representable as a positive integer, and so negating-2147483648produces2147483648, which produces an invalid sign-extended value when shifted. Sign-extension is only performed when targeting 64-bit platforms, where2147483648 >> 31evaluates to 1.This PR changes the sign-extension to no longer assume exactly 32-bit integers.