fix(checker): cap template literal type size to avoid unbounded growth

[Eugene Kalinin (ekalinin)]

Fixes #63271

The repro from the issue is a recursive conditional type whose check type becomes any after a few iterations:

type Dec<N extends number> =
    N extends 5 ? 4 : N extends 4 ? 3 : N extends 3 ? 2 :
    N extends 2 ? any :
    N extends 1 ? 0 : 0;

type Recur<N extends number, S extends string> =
    N extends 0 ? S : Recur<Dec<N>, `${S}_${S}`>;

type Explode = {
    [P in Recur<5, "a"> as `${P}_key`]: any;
};

With N = any, getConditionalType includes both branches and keeps tail-recursing with Dec<any> = any, while S doubles on every iteration. The tail recursion limit of 1000 iterations is never reached, because the string grows exponentially and memory runs out first.

The JS compiler threw RangeError: Invalid string length. The Go port has no string length limit, so it never terminates: after 45 seconds the process was at 9.8 GB RSS and still growing.

Fix

getTemplateLiteralType now bounds the size of the type it produces and reports TS2589 (Type instantiation is excessively deep and possibly infinite) when a limit is exceeded, returning errorType. This mirrors how checkCrossProductUnion bounds union sizes.

Two limits are needed:

• Text length (50,000,000 characters). This is the case from the issue.
• Number of placeholders (100,000). Once the length guard returns errorType, any is a valid placeholder, so ${any}_${any} starts doubling the number of spans instead of the text. The same growth happens in tsc.js for Recur<5, string>.

After the guard fires, the conditional loop continues cheaply on cached types until the existing tail recursion limit produces the final TS2589. Duplicate diagnostics at the same location are deduplicated, so the user sees a single error.

Choice of limits

The type-challenges test suite for "Length of String 3" (#31824) builds string literal types of up to ~10^7 characters and compiles today, so the length limit keeps a 5x margin over it. Going higher makes the regression test proportionally more expensive: at 100,000,000 it took 55 s and 2.2 GB under -race.

Test

templateLiteralTypeExcessiveLength.ts covers both growth patterns. Each reports exactly one TS2589.

Case	Wall time	Peak RSS
String doubling (issue repro)	~1 s	430 MB
Placeholder doubling	~0.7 s	182 MB

hereby test and hereby lint pass. No existing baselines changed.

[Copilot]

🟡 Changes recommended

The text-length guard can be bypassed by template literals split across multiple segments, leaving memory growth unbounded.

Once you've addressed the issues Copilot identified, you can request another Copilot review.

Pull request overview

Adds safeguards against unbounded template-literal type growth in the Go checker.

Changes:

• Caps template text length and placeholder count.
• Reports TS2589 when limits are exceeded.
• Adds regression tests and compiler baselines.

File summaries

File	Review
tsc/testdata/tests/cases/compiler/templateLiteralTypeExcessiveLength.ts	Adds regression cases.
tsc/testdata/baselines/reference/compiler/templateLiteralTypeExcessiveLength.types	Records type baseline.
tsc/testdata/baselines/reference/compiler/templateLiteralTypeExcessiveLength.symbols	Records symbol baseline.
tsc/testdata/baselines/reference/compiler/templateLiteralTypeExcessiveLength.js	Records emitted output.
tsc/testdata/baselines/reference/compiler/templateLiteralTypeExcessiveLength.errors.txt	Records expected diagnostics.
tsc/internal/checker/checker.go	Adds size guards. Critical (2 votes): The length check only measures the current builder segment, allowing aggregate text across reset segments to exceed the limit. Track cumulative segment length.

Review details

• Files reviewed: 6/6 changed files
• Comments generated: 1
• Review effort level: Balanced

---

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

[Copilot]

🟢 Approval recommended

All reviewed changes are covered by regression tests, with no unresolved issues.

Review details

• Files reviewed: 6/6 changed files
• Comments generated: 0 new
• Review effort level: Balanced