fix: indent standalone blocks in tseynit

[huytdps13400]

Summary

• indent standalone and nested block opening braces using the caller-provided indentation
• omit only the first indentation prefix when a statement is rendered inline after if/else/for/while/for-of
• avoid trimStart, so the stringifier never removes semantic or future formatter-owned whitespace
• keep the focused nested-block regression and affected obfuscation snapshots

Closes #2894.

Verification

• focused tseynit suite: 29/29 passed
• unplugin-typegpu obfuscation suite: 24/24 passed
• full source unit suite: 275 files; 2,853 passed, 2 skipped
• full monorepo typecheck: 0 errors
• full Oxlint/Prettier/Oxfmt style gate passed
• git diff --check passed

[Copilot]

Pull request overview

This PR fixes indentation behavior in the tseynit AST-to-JS stringifier so that standalone/nested blocks respect the caller-provided indentation, while control-flow statements (if/else/for/while/for-of) still keep their bodies directly after the control syntax.

Changes:

• Indent block opening braces using the provided indentation (fixes standalone/nested blocks starting at column 0).
• Ensure control-flow bodies don’t gain extra leading indentation by trimming the stringified body prefix.
• Add a targeted regression test and update affected obfuscation inline snapshots.

Reviewed changes

Copilot reviewed 2 out of 3 changed files in this pull request and generated no comments.

File	Description
packages/typegpu/src/shared/tseynit.ts	Adjusts stringifyStatement block formatting and control-flow body rendering to preserve correct indentation.
packages/typegpu/tests/internal/tseynit.test.ts	Adds a focused regression test covering standalone nested blocks + control-flow formatting expectations.
packages/unplugin-typegpu/test/obfuscation.test.ts	Updates obfuscation snapshots to match the corrected block indentation output.

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

[pullfrog]

✅ No new issues found.

Reviewed changes

• tseynit.ts — indent block opening braces. NODE.block now emits ${ident}{\n…\n${ident}}, so standalone blocks inherit the caller's indentation instead of starting at column 0.
• Control-flow bodies trimmed. if/else, for, while, and for-of bodies call .trimStart() so the block { still hugs the ) — exactly the double-space failure mode the issue (#2894) flagged.
• Regression + snapshot updates. New tseynit.test.ts case pins both cases (indented standalone block, inline if-body); the obfuscation snapshots reflect only the corrected { indent.

The fix matches the issue intent precisely, and the trimmed blast radius is worth noting: stringifyNode output feeds only error-message text (wgslGenerator.ts) and test snapshots — never emitted WGSL — so this is a formatting-only correction. I verified .trimStart() is safe across every statement type (each stringifies to a first line prefixed by exactly the indent), and nested standalone blocks inside a control-flow body still indent correctly.

  ^{｜ View workflow run ｜ Using DeepSeek Flash (free via Pullfrog for OSS) ｜ 𝕏}

[aleksanderkatan]

Can we do this without trim?

[huytdps13400]

Implemented without trimming in 214b4ebb.

stringifyStatement now accepts an omitInitialIndent flag. Control-flow bodies omit only their known first indentation prefix, while nested lines keep the caller-provided indentation unchanged. There is no trimStart left in the stringifier.

Focused verification: tseynit 29/29 and obfuscation 24/24. Full source verification: 275 files, 2,853 tests passed / 2 skipped; monorepo typecheck and full style gate pass.

[pullfrog]

✅ No new issues found.

Reviewed changes

• Replaced .trimStart() with a threaded omitInitialIndent parameter. stringifyStatement now takes a third argument and every statement type (expression, block, return, let, const, continue, break) returns via an initialIndent that is elided only for control-flow bodies (if/else, for, while, forOf). This is behaviorally byte-identical to the previously-approved .trimStart() version — the tseynit regression test and the obfuscation snapshots are unchanged in this commit.
• Cleaner than trimming. Omitting the caller-owned indent at the source instead of post-hoc .trimStart() means the stringifier never strips whitespace it didn't itself add, which is more robust against future formatter-owned or semantic leading space.
• Verified against the regression case. Traced if (true) { [3, 4]; } and the standalone nested { [1, 2]; } block through the new code path: the { of a control-flow body hugs the ) (no double space), while the standalone block opens indented and closes at ident. The inline snapshot matches exactly.

The refactor preserves the prior fix precisely while removing the whitespace-trim bluntness the earlier version carried.

  ^{｜ View workflow run ｜ Using DeepSeek Flash (free via Pullfrog for OSS) ｜ 𝕏}