mirror of
https://github.com/storytold/pdfcraft.git
synced 2026-11-10 13:01:29 +00:00
Grzegorz MartensandClaude Opus 5.5
fee25bfe4d
Reviews kept finding ways a small file could make this run long or large: readable() fetched the same objects again for each image that names a dictionary (~3.9 s each when two large objects push each other out of cos's caches), and the memo of names grew with distinct arrays times their entries (~3.7 GB from 2 MiB). Nothing bounded the run as a whole. Everything unshare reads now goes through one Run: what a reference leads to is fetched at most once a run for each way it is read (a size, a filter, parameters, an entry, a flag, /F, a colour space, an ICC profile), looked up before it is fetched and kept only as what is read of it. Objects fetched, names remembered, and what is decoded (a unit for each 4 KiB) count against RUN_BUDGET (2^22), and past it the run is refused with SharedJbig2Budget. A scan reads a few objects a page, so a book of thousands of pages sharing one dictionary stays well within it. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01AmsdeUCdTaD72jkDw2pJKY