Fix doc.text() throwing NaN with lineBreak false and underline/strike/link/goTo

[youdie006]

What kind of change does this PR introduce?

Bug fix. Fixes #401.

doc.text(str, x, y, { lineBreak: false, underline: true }) throws Error: unsupported number: NaN. The same happens with strike, link and goTo — anything that reaches _fragment's rendered-width calculation.

The chain: _initOptions only assigns result.width when result.lineBreak !== false (lib/mixins/text.js:421), and _text only constructs a LineWrapper if (options.width) (lib/mixins/text.js:82). But options.textWidth and options.wordCount are written only inside the wrapper (lib/line_wrapper.js:238-239), so with lineBreak: false they stay undefined and

const renderedWidth =
  options.textWidth + wordSpacing * (options.wordCount - 1) + characterSpacing * (text.length - 1);

evaluates to NaN. That reaches lineTo() and then PDFObject.number, whose range guard n > -1e21 && n < 1e21 is false for NaN, so it throws (lib/object.js:164-170).

The 2015 report describes this as invisible text rather than a crash, because number() did not throw back then. The title understates what it does today.

The fix supplies only the two missing inputs, leaving the wrapper's formula alone. _line 80 lines up already handles the no-wrapper case this way — this.x += this.widthOfString(text, options) (lib/mixins/text.js:449) — as does boundsOfString in its non-wrapped branch (lib/mixins/text.js:186). So the library already knows how to measure an unwrapped fragment; _fragment was just missing the fallback.

The values are local consts and are never assigned back to options, so continued-text state and _textOptions inheritance are untouched. I grepped the other readers of that state before changing it — line_wrapper.js:238-239 (the writer), line_wrapper.js:356, text.js:178-179 (inside a wrapper callback) and text.js:475 (guarded by if (options.width)) — none of them see a different value than before.

Verification. Built from master (c6f57c6) rather than testing the published package, since this area has moved recently.

Before the change, four public-API combinations throw and three controls pass:

FAIL  lineBreak:false + underline -> unsupported number: NaN
FAIL  lineBreak:false + strike    -> unsupported number: NaN
FAIL  lineBreak:false + link      -> unsupported number: NaN
FAIL  lineBreak:false + goTo      -> unsupported number: NaN
OK    underline alone / lineBreak:false alone / lineBreak:false + width + underline

After it, all seven pass. The four new cases in tests/unit/text.spec.js are verified red-green: with lib/mixins/text.js reverted and the tests kept, all four fail with unsupported number: NaN.

Full unit suite: 432 passing on unmodified master, 436 with this change — the delta is exactly the four new cases, with no existing test changing state. yarn lint is clean.

Checklist:

• Unit Tests
• Documentation — N/A, this is a crash fix with no public API or behaviour change for existing working input
• Update CHANGELOG.md
• Ready to be merged

Credit to @ashelley for the report.

Disclosure: I use AI assistance in my work, and I review and verify everything before it goes out. The runs above are my own.

[blikblum]

Many thanks