fix(printer): avoid double backslash in regex with escaped non-ASCII

[robobun]

Summary

• Fixes regex printing of backslash-escaped non-ASCII characters
• Tracks whether previous character was a backslash to avoid doubling backslashes when generating unicode escape sequences

Problem

When the regex printer encounters a non-ASCII character that was preceded by a backslash escape, it was incorrectly adding another backslash before the unicode escape sequence, resulting in \\uXXXX instead of \uXXXX.

For example:

const R = /[\⁄]/;  // backslash + U+2044 (fraction slash)

Was being printed as /[\\u2044]/, which changes the regex semantics:

• Expected: matches the fraction slash character (⁄)
• Actual (bug): matches \, u, 2, 0, 4, 4 as separate characters

This caused the regex from the issue to fail matching fractions:

const R = /[½⅓⅔¼¾⅕⅖⅗⅘⅙⅚⅐⅛⅜⅝⅞⅑⅒]|([⁰¹²³⁴⁵⁶⁷⁸⁹]+|[₀₁₂₃₄₅₆₇₈₉]+|[0-9]+)([\/\⁄])([⁰¹²³⁴⁵⁶⁷⁸⁹]+|[₀₁₂₃₄₅₆₇₈₉]+|[0-9]+)/;
'³⁄₅₂ cup of stuff'.match(R);
// Expected: ["³⁄₅₂", "³", "⁄", "₅₂"]
// Actual: null

Solution

Track whether the previous character was a backslash when iterating through the regex literal. When generating unicode escape sequences for non-ASCII characters, omit the leading backslash if the previous character was already a backslash (which was escaping the non-ASCII character).

Test plan

• Added regression test in test/regression/issue/26785.test.ts
• Test verifies the regex source is correctly printed
• Test verifies the regex matches the expected characters

Closes #26785

🤖 Generated with Claude Code

[robobun]

^{Updated 10:49 PM PT - Feb 6th, 2026}

❌ Your commit 6757c4d1 has 1 failures in Build #36670 (All Failures):

• test/integration/next-pages/test/dev-server.test.ts - 1 failing on 🍎 14 aarch64

---

🧪   To try this PR locally:

bunx bun-pr 26786

That installs a local version of the PR into your bun-26786 executable, so you can run:

bun-26786 --bun

[coderabbitai]

Walkthrough

Adds tracking of whether the previous character was a backslash during bun_platform non-ASCII escaping, and uses that to choose between emitting full Unicode escapes (\uXXXX) or partial escapes (uXXXX) in template literals and RegExp literals. Also adds regression tests for the behavior.

Changes

Cohort / File(s)	Summary
Non-ASCII Character Escaping Logic src/js_printer.zig	Adds prev_was_backslash state initialization and reset in the bun_platform non-ASCII escaping routine. Updates emission paths to choose between \\uXXXX and uXXXX for BMP code points and between paired \\uXXXX/uXXXX sequences for surrogate pairs. Mirrors logic in RegExp literal printing to avoid doubled backslashes.
Regression Test Suite test/regression/issue/26785.test.ts	Adds three Bun-based regression tests verifying printing and matching of backslash-escaped and non-escaped non-ASCII characters in regex literals (including fraction slash cases). Tests run via Bun.spawn and assert source, match contents, index, and exit code.

🚥 Pre-merge checks | ✅ 4

✅ Passed checks (4 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title accurately describes the main fix: avoiding double backslash in regex with escaped non-ASCII characters, which is the primary change in this PR.
Description check	✅ Passed	The description provides clear sections for problem, solution, and test plan, though it doesn't strictly follow the template structure with 'What does this PR do?' and 'How did you verify?' headings.
Linked Issues check	✅ Passed	The PR successfully addresses issue #26785 by fixing the regex printer to correctly handle backslash-escaped non-ASCII characters, preventing double backslashes and enabling proper regex matching.
Out of Scope Changes check	✅ Passed	All changes are directly related to fixing the backslash-unicode escape sequence issue in both the JS printer and adding regression tests, with no unrelated modifications detected.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

---

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

[coderabbitai]

Actionable comments posted: 1

🤖 Fix all issues with AI agents

In `@test/regression/issue/26785.test.ts`:
- Around line 11-33: Replace the tempDir + file I/O pattern used in the test
(the using dir = tempDir("issue-26785", { "test.js": ... }) and subsequent
Bun.spawn with cwd) with a single-file spawn using bunExe() -e and inline
source: stop creating "test.js" on disk, pass the script string as the argument
to bun via ["bunExe()", "-e", "<script source>"] in the Bun.spawn call, remove
cwd/file-related setup and teardown, and apply the same replacement to the other
two test cases referenced (the blocks around lines 47–62 and 78–97) so each test
uses bunExe() with -e and no tempDir.