fix(transforms/module): replace canonicalization with clean to not mess up symlinks by perbergland · Pull Request #11585 · swc-project/swc

perbergland · 2026-02-24T18:37:08Z

Summary

Replaces std::fs::canonicalize() with path_clean::PathClean::clean() in both NodeImportResolver::try_resolve_import and ModuleConfig::get_resolver
This normalizes . and .. path components without resolving symlinks
No new config options needed — this is a pure bugfix

Root cause

When jsc.baseUrl is set, SWC creates a NodeImportResolver which called std::fs::canonicalize() on resolved paths (path.rs:261). canonicalize() does three things: makes paths absolute, resolves ./.., and resolves symlinks. Only the first two are needed — the symlink resolution breaks setups where symlinked source files need imports resolved relative to the symlink location.

The path_clean crate (already a dependency) provides .clean() which does the same normalization lexically, without touching the filesystem or resolving symlinks.

Changes

crates/swc_ecma_transforms_module/src/path.rs: Replace canonicalize(resolved) with resolved.clean(), remove unused fs::canonicalize import
crates/swc/src/config/mod.rs: Replace v.canonicalize() with v.clean() (+ make-absolute for relative paths) in get_resolver() so both base and target use the same non-symlink-resolving normalization — this ensures diff_paths compares paths in the same "path space"
crates/swc/Cargo.toml: Add path-clean dependency
crates/swc_ecma_transforms_module/tests/path_node.rs: Add test with symlinked directory that verifies import paths are preserved through symlinks

Test plan

New test issue_11584_symlink_not_canonicalized creates a symlinked directory structure and verifies ../lib/dep resolves to ../lib/dep.js (not ../../real/lib/dep.js)
issue_8667_1 Bazel sandbox test passes (confirmed diff_paths produces correct relative paths with consistent normalization)
All 260 swc_ecma_transforms_module tests pass
All 9 swc_cli_impl tests pass
cargo clippy and cargo fmt clean

🤖 Generated with Claude Code

swc-project#11584) Add test that demonstrates the issue where `canonicalize()` in `NodeImportResolver::try_resolve_import` resolves symlinked paths to their real filesystem location, breaking import resolution when `preserve_symlinks` is intended. Also adds the `preserve_symlinks` field to `path::Config` and threads it through `JscConfig` and `build_resolver`, but does not yet skip the canonicalization (test is expected to fail). Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

…et (swc-project#11584) When `jsc.preserveSymlinks` is true, skip the `canonicalize()` call in `NodeImportResolver::try_resolve_import` so that symlinked paths are preserved during module resolution. Also skip base filename canonicalization in `ModuleConfig::get_resolver`. This fixes the issue where `jsc.baseUrl` causes symlinked source files to have their imports resolved from the real file location instead of the symlink location, breaking bundlers that set `resolve.symlinks: false`. Closes swc-project#11584 Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

CLAassistant · 2026-02-24T18:37:16Z

All committers have signed the CLA.

changeset-bot · 2026-02-24T18:37:17Z

🦋 Changeset detected

Latest commit: 9e5fc33

The changes in this PR will be included in the next version bump.

Not sure what this means? Click here to learn what changesets are.

Click here if you're a maintainer who wants to add another changeset to this PR

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

codspeed-hq · 2026-02-24T18:40:41Z

Merging this PR will not alter performance

✅ 192 untouched benchmarks

_{Comparing perbergland:fix/preserve-symlinks-11584 (9e5fc33) with main (86815b1)}

chatgpt-codex-connector

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 09f010577d

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

Open a pull request for review
Mark a draft as ready
Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

crates/swc/src/config/mod.rs

github-actions · 2026-02-24T18:49:12Z

Binary Sizes

File	Size
`swc.linux-x64-gnu.node`	28M (28745160 bytes)

Commit: dfbb790

…et (swc-project#11584) Replace `std::fs::canonicalize()` with `path_clean::PathClean::clean()` in `NodeImportResolver::try_resolve_import`. This normalizes `.` and `..` path components without resolving symlinks, fixing the issue where symlinked source files had their imports resolved from the real file location instead of the symlink location. `canonicalize()` was originally added for Bazel sandbox support (swc-project#8265), but its symlink resolution is unnecessary — lexical path cleaning achieves the same normalization without breaking symlinked source files. This removes the `preserve_symlinks` config option added in the previous commit since it is no longer needed. Closes swc-project#11584 Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

socket-security · 2026-02-25T08:04:00Z

No dependency changes detected. Learn more about Socket for GitHub.

👍 No dependency changes detected in pull request

chatgpt-codex-connector

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 5fb410a47d

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

Open a pull request for review
Mark a draft as ready
Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

crates/swc_ecma_transforms_module/src/path.rs

…roject#11584) Also normalize the base filename in `get_resolver()` using `path_clean::PathClean::clean()` instead of `std::fs::canonicalize()`. This ensures both base and target paths live in the same "path space" so that `diff_paths` computes correct relative paths in Bazel sandbox and symlinked setups. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

chatgpt-codex-connector

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 152d8171ae

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

Open a pull request for review
Mark a draft as ready
Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

crates/swc/src/config/mod.rs

) The `path_clean::PathClean` import is only used inside the `#[cfg(feature = "module")]` impl block. Gate the import to avoid unused-import errors when building without the module feature. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

kdy1

Not sure if this is what we want. I want to ensure that it works for bazel users before merging/publishing

kdy1 · 2026-03-03T08:16:18Z

I'm sorry to bother you all, but I want to ensure this doesn't break rules_swc.
Can you check if this PR looks good?

cc @alexeagle @gregmagolan @jbedard

jbedard · 2026-03-03T20:50:53Z

Thanks for the ping @kdy1. I think this would be a positive change for rules_swc/js, normally bazel issues come from tools that do resolve or follow symlinks.

Have you observed any issues with things like pnpm package stores? I assume import '@a/pkg is not effected by this change, but what about import './node_modules/@a/pkg? If that is a symlink into ./node_modules/.pnpm/node_modules/@a/pkg I'm wondering if that is an example that this PR fixes?

kdy1 · 2026-03-03T23:33:30Z

@claude Add tests for the use case described in #11585 (comment)

claude · 2026-03-03T23:33:48Z

Claude Code is working…

I'll analyze this and get back to you.

View job run

chatgpt-codex-connector

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 3c0467f370

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

Open a pull request for review
Mark a draft as ready
Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

chatgpt-codex-connector · 2026-03-07T00:47:14Z

crates/swc/src/config/mod.rs

+                    env::current_dir()
+                        .map(|cwd| cwd.join(v).clean())
+                        .unwrap_or_else(|_| v.to_path_buf())


Resolve missing relative base against baseUrl

When base is relative, this new logic always makes it absolute via env::current_dir(). In in-memory/virtual-file compiles (relative filename that does not exist on disk), the previous behavior kept base relative, so NodeImportResolver::try_resolve_import could rebase it through absolute_path(base_dir, ...) and keep rewrites in the jsc.baseUrl path space; with this change, base is pinned to process CWD instead, and diff_paths can emit incorrect long paths (for example ../../project/...) when cwd differs from baseUrl.

Useful? React with 👍 / 👎.

chatgpt-codex-connector

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 18cf53e24b

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

Open a pull request for review
Mark a draft as ready
Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

chatgpt-codex-connector · 2026-03-07T01:55:50Z

crates/swc/src/config/mod.rs

-            FileName::Real(v) if !skip_resolver => {
-                FileName::Real(v.canonicalize().unwrap_or_else(|_| v.to_path_buf()))
-            }
+            FileName::Real(v) if !skip_resolver => FileName::Real(v.clean()),


Keep Windows base path in UNC form before diffing

Replacing canonicalize() with clean() here leaves base as a drive-letter path on Windows, but build_resolver still canonicalizes jsc.baseUrl to UNC (\\?\...) before creating NodeImportResolver; that means try_resolve_import can compare two absolute paths with different prefixes and skip normalization (it only normalizes when absolute-vs-relative differs), so relative rewriting via diff_paths no longer works and imports can fall back to absolute //?/C:/... specifiers for jsc.baseUrl/paths resolutions.

Useful? React with 👍 / 👎.

chatgpt-codex-connector

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 9e5fc339f7

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

Open a pull request for review
Mark a draft as ready
Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

chatgpt-codex-connector · 2026-03-07T02:27:53Z

crates/swc_ecma_transforms_module/src/path.rs

-            }
+        // https://github.com/swc-project/swc/issues/11584
+        if let FileName::Real(resolved) = &mut target.filename {
+            *resolved = resolved.clean();


Normalize Windows target paths before import diffing

On Windows, replacing canonicalization with resolved.clean() leaves absolute targets in drive-letter form (for example C:\...), while resolver setup still canonicalizes jsc.baseUrl to UNC (\\?\C:\...) in build_resolver; TsConfigResolver can also return absolute FileName::Real(tp.into()) directly for exact paths mappings (crates/swc_ecma_loader/src/resolvers/tsc.rs, Pattern::Exact). In that case diff_paths receives two absolute paths with different prefixes, fails to compute a relative path, and the rewriter falls back to absolute specifiers, which is a regression for Windows projects using absolute path mappings.

Useful? React with 👍 / 👎.

perbergland and others added 2 commits February 24, 2026 19:32

perbergland requested review from a team as code owners February 24, 2026 18:37

chore: add changeset for preserveSymlinks option

e070d00

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

chatgpt-codex-connector bot reviewed Feb 24, 2026

View reviewed changes

crates/swc/src/config/mod.rs Outdated Show resolved Hide resolved

perbergland mentioned this pull request Feb 25, 2026

3.4 modern build rspack doesn’t support symlinked source files meteor/meteor#14053

Open

chatgpt-codex-connector bot reviewed Feb 25, 2026

View reviewed changes

crates/swc_ecma_transforms_module/src/path.rs Show resolved Hide resolved

perbergland changed the title ~~fix(transforms/module): add preserveSymlinks option to skip canonicalization~~ fix(transforms/module): replace canonicalization with clean to not mess up symlinks Feb 25, 2026

chatgpt-codex-connector bot reviewed Feb 25, 2026

View reviewed changes

crates/swc/src/config/mod.rs Show resolved Hide resolved

kdy1 reviewed Mar 3, 2026

View reviewed changes

test(transforms/module): add pnpm symlink resolver coverage

3c0467f

chatgpt-codex-connector bot reviewed Mar 7, 2026

View reviewed changes

kdy1 added 2 commits March 7, 2026 09:48

Update fix-preserve-symlinks.md

f6353f5

fix(swc): keep relative resolver base in baseUrl space

18cf53e

chatgpt-codex-connector bot reviewed Mar 7, 2026

View reviewed changes

kdy1 added 2 commits March 7, 2026 11:00

fix(swc): keep windows resolver base in UNC form

d81aaf3

fix(swc): normalize existing relative filenames in cwd space

9e5fc33

chatgpt-codex-connector bot reviewed Mar 7, 2026

View reviewed changes

Uh oh!

Conversation

perbergland commented Feb 24, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Summary

Root cause

Changes

Test plan

Uh oh!

CLAassistant commented Feb 24, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

changeset-bot bot commented Feb 24, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

🦋 Changeset detected

Uh oh!

codspeed-hq bot commented Feb 24, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Merging this PR will not alter performance

Uh oh!

chatgpt-codex-connector bot left a comment

Choose a reason for hiding this comment

💡 Codex Review

Uh oh!

Uh oh!

github-actions bot commented Feb 24, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Binary Sizes

Uh oh!

socket-security bot commented Feb 25, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

chatgpt-codex-connector bot left a comment

Choose a reason for hiding this comment

💡 Codex Review

Uh oh!

Uh oh!

chatgpt-codex-connector bot left a comment

Choose a reason for hiding this comment

💡 Codex Review

Uh oh!

Uh oh!

kdy1 left a comment

Choose a reason for hiding this comment

Uh oh!

kdy1 commented Mar 3, 2026

Uh oh!

jbedard commented Mar 3, 2026

Uh oh!

kdy1 commented Mar 3, 2026

Uh oh!

claude bot commented Mar 3, 2026

Uh oh!

chatgpt-codex-connector bot left a comment

Choose a reason for hiding this comment

💡 Codex Review

Uh oh!

chatgpt-codex-connector bot Mar 7, 2026

Choose a reason for hiding this comment

Uh oh!

chatgpt-codex-connector bot left a comment

Choose a reason for hiding this comment

💡 Codex Review

Uh oh!

chatgpt-codex-connector bot Mar 7, 2026

Choose a reason for hiding this comment

Uh oh!

chatgpt-codex-connector bot left a comment

Choose a reason for hiding this comment

💡 Codex Review

Uh oh!

chatgpt-codex-connector bot Mar 7, 2026

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Milestone

Development

Uh oh!

perbergland commented Feb 24, 2026 •

edited

Loading

CLAassistant commented Feb 24, 2026 •

edited

Loading

changeset-bot bot commented Feb 24, 2026 •

edited

Loading

codspeed-hq bot commented Feb 24, 2026 •

edited

Loading

github-actions bot commented Feb 24, 2026 •

edited

Loading

socket-security bot commented Feb 25, 2026 •

edited

Loading