codegen: fix segfault on numeric literal assigned to a pointer type - #22
Merged
Merged
Conversation
Hedede
force-pushed
the
test/gen-local-null-store
branch
from
July 16, 2026 13:48
6fbfd93 to
6cd11d8
Compare
`var p: int* = 0;` segfaults the compiler. Root cause: gen(numeric_literal) falls through to `return nullptr` when the literal's inferred type doesn't match integer/float/double (a plain int literal used to initialize a pointer-typed local isn't handled), and gen_local() passes that nullptr straight into builder.CreateStore() without checking it, which dereferences it and crashes. Confirmed via gdb: SIGSEGV inside backend_llvm::gen_local, called from gen(decl_statement) -> gen(statement_block) -> gen(function) -- the same call chain that flows into the "TODO: this bit crashes sometimes" comment a few frames up in gen(function); this is a concrete, 100% reproducible instance of that class of bug (a null Value* from gen() reaching a raw LLVM builder call unchecked), not a fix.
Root cause of the crash reproduced by codegen/pointer_null_init.aw: propagate_type(numeric_literal) would blindly assign any target type (including a pointer type) onto a literal's expr.type with no check, and gen(numeric_literal) had no case for a pointer-typed literal, so it fell through to `return nullptr`. That nullptr then flowed unchecked into gen_local()'s builder.CreateStore(), which dereferences it immediately and segfaults. Fix: - type_inference: allow only a literal `0` to propagate a pointer type (the null-pointer idiom); anything else assigned to a pointer type is now a clean "Type mismatch" diagnostic instead of silently producing bad IR (or worse, corrupting the compiler's own state). - codegen: gen(numeric_literal) now emits ConstantPointerNull for a pointer-typed literal, so `var p: T* = 0;` generates correct IR. - gen_local(): check gen(var.value) for null before handing it to CreateStore(), so any future unhandled case fails cleanly instead of segfaulting -- this is the same unchecked-null pattern flagged by the "TODO: this bit crashes sometimes" comment in gen(function). Added test/errors/pointer_nonzero_init.aw as a negative test for the new diagnostic. codegen/pointer_null_init.aw now passes cleanly.
The crash is fixed, so the test can be re-enabled
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
2 participants
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.
Summary
var p: int* = 0;reliably segfaulted the compiler. Root cause:propagate_type(numeric_literal)assigned any target type — including a pointer type — onto a literal with no validation, andgen(numeric_literal)had no case for pointer types, falling through toreturn nullptr. That nullptr then flowed unchecked intogen_local()'sbuilder.CreateStore(), which dereferences it immediately.type_inference.c++: only a literal0may now propagate a pointer type (the null-pointer idiom); anything else assigned to a pointer type produces a cleanType mismatch: '<type>' vs 'numeric_literal'.diagnostic instead of crashing or silently emitting bad IR.backend_llvm.c++(gen(numeric_literal)): pointer-typed literals now emitConstantPointerNull, sovar p: T* = 0;generates correct IR (store ptr null, ...).backend_llvm.c++(gen_local): added a null check beforeCreateStore, so any future unhandled case fails cleanly instead of segfaulting — the same unchecked-null pattern flagged by the pre-existing// TODO: this bit crashes sometimescomment ingen(function).Test plan
SIGSEGVinsidebackend_llvm::gen_local, called viagen(decl_statement) → gen(statement_block) → gen(function).test/codegen/pointer_null_init.aw(previously segfaulting, now compiles cleanly and passes) andtest/errors/pointer_nonzero_init.aw(negative test for the new diagnostic).ctest: 19/19 pass, excluding the pre-existing, unrelatedcodegen/pointer.awfailure.Generated by Claude Code