mirror of
https://github.com/storytold/pdfcraft.git
synced 2026-11-10 13:01:29 +00:00
Grzegorz MartensandClaude Opus 5.5
8bf63c8792
The Type 3 cleanup at the end of a redaction (unlink_type3) scans every page and every form it reaches for the glyphs they show, and reads the programs of the glyphs it keeps for the resources they use. Both parsed each stream whole before checking for bytes they couldn't read, outside the page's allowance and up to 1 GiB a stream. So a form with resources of its own outside the mark, which no other reader follows, cost about 79 bytes of memory per byte of content: the peak grew by 632 MiB for 8 MiB of `n` operators. On main too, wherever a Type 3 font survives in a page's or a form's resources. Both now read with pdfcraft_content::ops, one operator at a time, and refuse for unreadable bytes once they have read the whole stream, as before. The scan only reads, and either error refuses the whole operation, so acting on operators before that check changes nothing. The memory test gains a case for each: the_forms_followed_through_resources_are_read_one_operator_at_a_time fails with either read reverted. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01AmsdeUCdTaD72jkDw2pJKY