carlobeltrame / ecamp3 / 30009113026

34%

devel: 43%

push

github

fix(print): drain PDF stream synchronously instead of via setTimeout(0) events

Eliminate the Firefox worker hang at its source instead of mitigating it.

@react-pdf/render renders synchronously and calls ctx.end() before returning,
and with compression disabled pdfkit writes the entire document into the stream
buffer synchronously. The bytes are therefore already available when we get the
instance, so toBlob() now drains them with read() rather than listening for
'data'/'end'. pdfkit emits those events through its bundled process/browser
nextTick polyfill (setTimeout(0)), which starves in Firefox web workers and
leaves toBlob() hanging forever (flaky in CI, where every render runs in a fresh
worker). read() needs none of that scheduling.

A streamToBlob unit test renders a 20-page pdfkit document and asserts the
read() drain yields a byte-identical, valid PDF (%PDF-...%%EOF) to the event
drain, guarding against silent truncation.

This supersedes the per-worker warm-up render (removed here); with the stream
no longer drained through the nextTick scheduler there is nothing to warm.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01KQ8DeKNRbyq2fLVAqdJfof

671 of 4489 branches covered (14.95%)

Branch coverage included in aggregate %.

4444 of 10648 relevant lines covered (41.74%)

63.95 hits per line