You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
(OPT) Compare canonical function-instance identities without temporary encoded buffers #853
Source-bound at 17b12c909916ce8671b43f208b59f3432144709a: src/hir/validation.rs:1264–1268 derives an owned FunctionInstanceId only to compare it with a retained identity. src/hir/ids.rs:578–592 builds both an encoded-argument CappedString and a final formatted owned ID. The same comparison pattern occurs in src/hir/validation/generic_template.rs:204,501,576,593. Existing ids.rs:600–638 has a borrowed exact-encoder precedent. Main b80 does not already change these files.
Implement an exact streaming canonical-ID comparison for these validation-only sites, preserving the complete allocating derive constructor as an independent oracle and construction API. Initially retain every argument.identity_key() producer; claim removal only of the actual encoded-argument buffer and final temporary ID. Fund genuinely new controls before their formation, or eliminate the extra carriers lexically. No cap, forecast, refund, guessed fee or diagnostic-order change.
Preserve exact grammar, UTF-8 and byte-length handling, argument ordering, forged/truncated/prefix IDs, duplicate insertion before canonical-identity refusal, fresh/sticky/floor failures and complete Drop. Exercise the existing harness with minimal direct success/negative and independent old/new controls after the implementation batch; scoped Clippy and actual moved pins only when needed. No full suite.
Historical 47f Catalog23 and original Shift schedule fail in HIR final validation at their unchanged 64 MiB ceiling. Their stage-only receipts do not establish that this operation selected the refusal, and no numerical saving or application fit is promised. Keep those original owners unchanged and use their later actual outcomes to judge fit. Work is being implemented in an isolated next-batch lane; no new runtime or agent-token measurement is claimed.
Activity