[anubhavlal07]

Purpose / Goal

Fixes #924.

A quote inside a processing instruction makes the parser skip past the real ?>. In <r><?note don't forget?><a>1</a><b>it's</b><?end?><c>3</c></r>, a, b and ?end are dropped. If there is no later matching quote, the parser throws Pi Tag is not closed.

Root cause: the PI branch in OrderedObjParser.js calls readTagExp(xmlData, i, false, "?>"), which calls tagExpWithClosingIndex. That function treats every ' or " as the start of an attribute value and jumps to the next matching quote. PI content is free text, and XML 1.0 production [16] ends a PI at the first ?>.

Fix: tagExpWithClosingIndex gets a skipQuotes parameter (default true). readTagExp passes closingChar !== "?>", so in PIs it does not skip quoted values or replace tabs. Tabs inside quoted PI pseudo-attribute values stay as they are. PI pseudo-attributes such as <?xml version="1.0"?> and xml-stylesheet are still parsed by buildAttributesMap on the extracted expression. Its regex already splits on any whitespace. Normal tags behave exactly as before.

Tests added to spec/pi_spec.js:

• an apostrophe in PI content keeps the elements that follow it
• an unmatched quote in PI content no longer throws
• a double quote in PI content (preserveOrder)
• a PI ends at the first ?> even inside a quoted value, which matches the validator's existing expectation
• tabs inside quoted PI attribute values are kept, which guards against a side effect of the change

The first four fail on master and pass with this change. Full suite: 348 specs, 0 failures.

Benchmark (node benchmark/XmlParser.mjs, spec/assets/ptest.xml). Runs on this machine vary a lot between identical runs, but there is no regression. The change only touches PI scanning.

	before	after
fxp	9259 req/s	9309 req/s
fxp - preserve order	12325 req/s	16821 req/s

Type

• Bug Fix
• Refactoring / Technology upgrade
• New Feature