postprocess: Preprocess the C source code passed to transforms#1846
Merged
Conversation
thedataking
force-pushed
the
perl/postprocess-json-c
branch
from
332513d to
c1d7cd9
Compare
thedataking
force-pushed
the
perl/postprocess-json-c-gpt-5.5-2026-04-23
branch
2 times, most recently
from
19142b4 to
24d771a
Compare
thedataking
force-pushed
the
perl/postprocess-json-c-gpt-5.5-2026-04-23
branch
from
24d771a to
e1ff226
Compare
thedataking
force-pushed
the
perl/postprocess-json-c-gpt-5.5-2026-04-23
branch
3 times, most recently
from
8ee5007 to
9d8d679
Compare
thedataking
force-pushed
the
perl/postprocess-json-c-gpt-5.5-2026-04-23
branch
from
9d8d679 to
9a5f17f
Compare
thedataking
requested review from
fw-immunant
and removed request for
Crocodoctopus
fw-immunant
approved these changes
Jun 17, 2026
fw-immunant
left a comment
fw-immunant
left a comment
Contributor
There was a problem hiding this comment.
Generally looks good; inline comments about a couple edge-cases and formatting in tests.
One larger question I have is how this interacts with configurability--is this an improvement for only the case where we fully remove all preprocessor logic (as opposed to lowering it to #[cfg]s with Hayroll?
The transpiler now emits an allow for clippy::missing_safety_doc, which shifts the expected function spans by one line.
Contributor
Author
Yes, I think it is the right for comment transfer today since it lets us validate transferred comments in more cases. I haven't thought much about how this would work when lowering |
thedataking
force-pushed
the
perl/postprocess-json-c-gpt-5.5-2026-04-23
branch
2 times, most recently
from
267c551 to
e0bbf48
Compare
Restructure *.c_decls.json from a flat identifier-to-snippet map into {definitions: {ident: {definition, preprocessed_definition}}}.
The AST exporter runs an in-process preprocessor-only pass per translation unit (clang -E -fdirectives-only -C equivalent): conditional directives are resolved, dropping inactive regions and their comments, while macro invocations and comments are preserved. Line markers map the output back to each function definition's source range; the main file is recognized by the spelling in the leading line marker, since marker paths are relative to the compilation directory. Functions whose mapping fails get a null preprocessed_definition.
Preprocessed snippets are keyed by C declaration id to avoid attaching a function body to another declaration with the same C spelling. Translation units with #line directives or non-monotonic main-file line markers skip preprocessed metadata and fall back to raw definitions. Clang versions without directives-only preprocessing emit a warning before using the legacy full-expansion path.
…prompts Read both forms from the structured *.c_decls.json instead of running a clang subprocess per snippet. The CommentsTransform prompt includes the preprocessed text only when it differs from the original, so prompts (and llm-cache keys) are unchanged for directive-free functions. Comment gating and response validation use comments from the preprocessed text plus comments on original preprocessor directive lines, so inactive-region comments are not transferred while directive-line comments are not dropped just because clang omits them from directives-only output.
Use a slotted dataclass for CommentsTransformPrompt and centralize the preprocessed-if-changed rule on CDefinition. This keeps prompt construction behavior unchanged while removing duplicated field and invariant handling.
thedataking
force-pushed
the
perl/postprocess-json-c-gpt-5.5-2026-04-23
branch
from
e0bbf48 to
0fa5b6c
Compare
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.
The
CommentsTransformcan fail if presented with a C and Rust function pair where some comments are inside directives that are compiled out during preprocessing and thus should not be transferred to the Rust.We can simply run the C compiler on the C snippet and it will pick up compile commands (if any) and give us the preprocessed version back.