Conversation
…26201) Sibling modules are compiled by separate, possibly parallel `nim m` processes, so an in-memory counter let both allocate from the same initial state and embed duplicate values. The counters now live in `<nimcache>/ic.counters`, guarded by an OS file lock that a process holds from its first counter operation until it is done with its module. The file records, per counter, the high-water mark and the numbers each module was handed: a re-semmed module gets the same numbers again (no-op rebuilds stay content-stable), a module needing more gets fresh ones above the high-water mark and never collides with values cached modules embed. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Field uses resolve against their owning type via one `loadFieldUse` (a local `lookupInRecord`/`lookupFieldAgain`, since ast2nif cannot import astalgo/types) and fall back to the stub instead of raising. The field of a constructor's `f: v` now resolves against the constructor's type too, so `nkObjConstr`/`nkTupleConstr` share PSym identity with the reclist. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
`writeBackendActions`/`applyBackendActions` take the module's `.c` path and derive the `.cflags` sidecar name themselves, so `localpassc` needs no path field at all: it applies to the C file the sidecar belongs to. Removes the suffix-stripping of the sidecar path. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Emit the wrapper for every type definition, even without sons, instead of only for six structural kinds: 4 bytes per type give every BIF consumer an unambiguous arity. The loader now requires the section; the format stamp already rules out older caches, so the raw-tail fallback is gone. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- sigmatch: a converter returning `var T` from a loaded module flagged its
sealed return type with tfVarIsPtr; flag a private copy instead.
- transf: treat nkReplayAction as a leaf; its payload (e.g. a macrocache
value holding an untyped proc AST) is data, not code.
- cgen: a routine owned by a module with `{.localPassC: "-m...".}` that is
emitted into another module's TU (generic instance, emit-everywhere copy)
is compiled for the owner's target via `#pragma GCC target` (clang:
`#pragma clang attribute`). ast2nif.loadedReplayActions reads the owner's
replay actions lazily.
- vm: getImpl of a loaded routine rebuilds its params as symbols, not
identifiers, so macros can query their types (shady's toShader).
- icconfig: the precompiled config now carries lazyPaths/nimblePaths, so
`nim ic` resolves nimble packages (IcConfigVersion 3).
- ast: skipGenericOwner tolerates a backend-lifted hook without owner.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The iface cookie hashes signatures only, so a body edit re-sems no importer. But the backend inlines foreign bodies into other modules: an inline proc into every using TU (`cg`), an inline iterator into the using module's lowered NIF (`lower`). Those rules listed only their own module's NIF, so the users kept the old body and `nim ic` diverged from `nim c`. Each `lower`/`cg` run now records the modules whose routine bodies it read (`ModuleGraph.icBodyDeps` via `getBody`; `BModule.icImplMods`, now including inline procs) in a `.bodydeps` sidecar, and deps.nim lists their `.s.bif` resp. `.t.bif` as inputs of the rule on the next run, like a `-MD` depfile. The metamorphic harness treats the backend build file as provenance, like the frontend one: it carries the discovered dependencies a clean build has not seen yet. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
This branch has not been deployed
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.
…(#26201)
Sibling modules are compiled by separate, possibly parallel
nim mprocesses, so an in-memory counter let both allocate from the same initial state and embed duplicate values. The counters now live in<nimcache>/ic.counters, guarded by an OS file lock that a process holds from its first counter operation until it is done with its module.The file records, per counter, the high-water mark and the numbers each module was handed: a re-semmed module gets the same numbers again (no-op rebuilds stay content-stable), a module needing more gets fresh ones above the high-water mark and never collides with values cached modules embed.