M3: carry chosen tools through native preparation and password retries, defer startup intake during pending opens, and await actual native completion in tests.
A stamp placed on a page with /Rotate was drawn in user space, so the
label or picture turned with the page. Counterrotate the appearance so
it reads upright as the page is displayed.
Co-authored-by: Brandon Thomas <echelon@gmail.com>
- Create PDF detects BigTIFF (II+/MM+) as well as classic TIFF; the tiff
decoder already reads it, so these scans now convert instead of being
refused.
- Image detection accepts only GIF87a/GIF89a, the headers conversion
decodes, so a malformed GIF header is no longer called an image and then
rejected with a misleading error.
- file-format-parity.md: XDP split into XFA data import/export (shipped
under form.xfa-data) and opening .xdp as a document (none).
Tests: bigtiff_images_become_pages; malformed GIF headers in
source_kind_tells_pdfs_images_and_text_apart. Both fail without the fix.
Claude-Session: https://claude.ai/code/session_01F8DFYWkDyTnEcin9PNkUW5
Co-authored-by: Claude <noreply@anthropic.com>
Co-authored-by: Brandon Thomas <echelon@gmail.com>
`set_info` changed only `/Info`. Files that carry XMP (LibreOffice and Word
exports, PDF/A) kept the old dc:title and dc:creator, so viewers that read
XMP first kept showing the old title, and PDF/A preflight reported the two
as different (6.6.3).
The matching XMP property (Title dc:title, Author dc:creator, Subject
dc:description, Keywords pdf:Keywords, Creator xmp:CreatorTool, Producer
pdf:Producer) is now replaced where it stands, added in a description of
its own, or removed with an empty value. The packet is edited as text and
the rest of it stays byte for byte; files without XMP get none, and a
packet that isn't UTF-8 is left as it was.
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* M8: redaction unlinks removed images and forms from inherited and shared resources
Apply Redactions and Remove Hidden Information give a rewritten page its own
pruned /Resources, but the dictionary the page drew through before could still
name the image or form that was removed:
- a /Resources inherited from a /Pages node;
- one resource dictionary object shared by several pages;
- the page's own dictionary, when a form without /Resources drew the image
through it.
A full-rewrite save keeps whatever is still referenced, so the removed image
samples and form text stayed in the saved file while the operation reported
success. Per-page resources (the case fixed in #1099) were not affected.
The interpreter now records the original object behind every XObject it drops
or replaces. After the page loop, each page-tree resource dictionary that still
names one loses that entry unless a page drawing through that dictionary still
draws the name (the same image on another page, or again outside a mark). A
dictionary that loses an entry is replaced by a private copy on its node, so a
dictionary object shared with other nodes is left alone for them. If a page's
use of a name cannot be told (unreadable content, a form that will not decode),
the operation fails closed.
Tests: redacted_xobjects_do_not_survive_through_inherited_or_shared_resources,
an_image_another_page_still_draws_through_the_same_resources_is_kept,
an_image_a_form_drew_through_the_pages_resources_does_not_survive,
a_hidden_layers_image_does_not_survive_through_inherited_resources.
Not covered here: originals named by a form XObject's own /Resources, by an
annotation appearance or by a pattern.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* M9: a cross-reference entry the reader sets aside is never an unchanged signed object
The reader that judges post-signing changes and the renderer that draws the
page resolve some cross-reference entries differently. A revision appended
after signing could redefine a page's content stream and give the object an
entry the reader sets aside while the renderer follows it:
- an in-use entry at offset 0 (the reader skips it; the renderer fails to read
the object there, rebuilds its index and takes the last definition);
- a second entry for the same object in one section (the reader keeps the
first, the renderer the last).
The classifier compared the signed object with itself, found nothing changed
and reported the signature Valid while the page showed the new content.
pdfcraft-cos now records the entries it does not take at face value
(Document::xref_oddities: offset-0 in-use entries other than object 0,
duplicate entries that differ, reserved cross-reference stream types, and a
hybrid /XRefStm entry that disagrees with the table), and can say whether an
entry locates the object itself (Document::xref_entry_is_exact). What it
resolves is unchanged.
classify_changes treats any such entry added after the signed revision as a
change that is never permitted (page content when the object was a signed page
or its content, other changes otherwise). An object whose content compares
equal at a new place is unchanged only when its entry is exact: where the
reader had to search for the object, another reader may arrive elsewhere.
Tests: a_cross_reference_entry_the_reader_sets_aside_does_not_hide_new_page_content,
a_cross_reference_entry_that_misses_its_object_is_not_an_unchanged_object,
cross_reference_entries_that_are_set_aside_are_recorded,
an_entry_that_misses_its_object_is_found_by_searching_but_is_not_exact.
This closes the two demonstrated triggers and their near variants. It does not
prove that the two parsers agree on every file: other differences in what each
tolerates may remain.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* M8: Apply Redactions is refused when the page still shows text under a mark
Redaction decides what lies under a mark from the positions its own
interpreter computes, and its verify pass reuses them. Where those differ from
the renderer's, a mark drawn over the text as displayed lies over glyphs the
interpreter places elsewhere, or does not see at all: they stayed in the file
and the operation reported success.
The case the test pins: a font and size set through an ExtGState /Font entry
(`gs`), which the renderer honours and the interpreter does not model. The same
kind of gap was observed for a font a form XObject takes from its caller, for
some predefined CMaps and for a rotated Type 3 font matrix. The standard-14
width error that first showed it is fixed at its root by d0d1ad81.
The engine now asks the renderer, which is what the user saw when placing the
marks. After redaction runs, the result is written as a save would write it
(without the marks' own overlay), its text is extracted, and a glyph still at
least half inside a mark fails the edit with the existing "could not be
verified" error. The edit runs on a copy, so the document is unchanged. A page
whose text cannot be read back fails the same way.
This makes such mismatches fail closed; it does not make those documents
redactable. Teaching the interpreter each of them is separate work.
Test: redaction_is_refused_when_the_page_still_shows_text_under_a_mark (the
line whose font is set by `gs` is refused and left unchanged; the same line set
with `Tf` is redacted).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* fix(edit): move existing text by its frame, and selections with the arrow keys (fixes#1520)
* docs(parity): arrow-key nudging and the paragraph frame are no longer gaps (#1520)
Two follow-ups the maintainer asked for after #1355.
The script session's worker now says when the engine is ready for the
script it was sent (its JavaScript context built; FormCalc at once),
and the script's time limit runs from then, with the engine's own
start limit before it, as the one-shot path always did: building a
context is not charged to the script, which matters on a busy machine.
A field inside an area shares its data node with a field of the same
name outside it (areas are left out of data paths). The in-place
refresh of the live form now follows the data path through the tree
and gives every field bound to the node the value the tree builder
would, each by its own template field (an exclusion group there means
a rebuild), and the engine's value setter, with an index of widgets by
data node, gives every widget bound to the node the value by its own
kind, hidden fields included, writing the data whenever the node's
text differs. The script's time now also starts after its host and
globals are set up. A stale parity note (choosing a picture for an
image field, delivered in #1175) is removed.
* fix(audit): a rectangle drawn as an m/l path is a cover like re (fixes#1510)
* docs(parity): a rectangle drawn as an m/l path counts as a cover (#1510)
Two gaps in code-to-Unicode, both found by probing the current
behaviour rather than by reading the code.
A font with no /Widths and no /ToUnicode fell back to StandardEncoding
whatever it was, so /Symbol code 0x61 -- `alpha` in Symbol's own
built-in encoding -- decoded as "a". ZapfDingbats and MacExpert were
read as Latin the same way. That map is what `redact` matches patterns
against, so a wrong character there is a silent redaction failure, and
a wrong character is worse than no character. Symbol, ZapfDingbats and
MacExpert now yield no mapping at all unless the font names a base
encoding, which is always honoured as the deliberate statement it is.
This keys on the base font *name*, not the /FontDescriptor symbolic
flag. A great many ordinary Latin TrueType subsets set that flag and
still rely on a Latin encoding, so suppressing on the flag would lose
real text across a large share of real documents. A test pins that
restraint so it is not "fixed" into a regression later.
Glyph names now follow the published naming algorithm instead of two
of its cases: a name is cut at its first `.` (`a.sc` is `a`), split on
`_` into components (`f_i` is "fi"), and `uniXXXX` carries any number
of four-digit values (`uni00660069` is "fi"). Names are resolved whole
or not at all, so half a ligature is never returned. `glyph_text`
returns a String for this; `glyph_unicode` stays as the single-character
view. Hex parsing rejects the leading `+` that `from_str_radix` would
otherwise accept, and a lone surrogate half.
Both halves were verified by reverting each fix and watching the
matching tests fail. The parity note on protect.redact-patterns now
says Symbol-encoded text is skipped rather than mis-decoded.
Still open, and an owner decision: Symbol and ZapfDingbats could be
mapped properly with a glyph-name-to-Unicode table, but the vendored
Adobe Glyph List is not on the AGENTS.md 1.1 closed list.
Co-Authored-By: Claude <noreply@anthropic.com>
* M12: preview text structure proposals for auto-tagging
* Auto-tag preview: no unregistered command, allowed on copy-restricted files like a11y.check
The preview is read-only with no UI command yet, like render_preview, so it no
longer names the Planned a11y.autotag catalog id (tool_table_is_well_formed).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* M12: preview text structure proposals for auto-tagging
* Keep read-only auto-tag preview separate from unavailable command
---------
Co-authored-by: Brandon Thomas <echelon@gmail.com>
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* M10: edit OCR text on scanned pages so the change shows
Editing a recognised word on a scan rewrote it as invisible OCR text, so
the page still showed the old word, and the whole OCR layer grouped into
one paragraph that rewrapped when one word changed.
- An edited invisible paragraph is drawn visibly, over a cover in the
paper colour sampled around it, in the scan's sampled ink colour
(or the chosen colour).
- OCR words group into phrases: a table cell or a tab stop splits them
and each scanned line stays apart; a phrase grows rightwards.
- Styled rewrites measure with the text matrix's horizontal scale, so
stretched OCR words keep their widths instead of overlapping.
- The inline editor covers the scanned word and grows like the result.
* M10: size OCR text from each word's letters
A word box from recognition is tight around the ink, but the text layer
assumed every box spans Helvetica's ascender-descender height. A
capitals-only word such as "DT" spans only the cap height, so its text
was set about 23% too small and its baseline floated up; editing it
then drew the new word visibly smaller than the scanned one.
The em and baseline now come from the word's letters: cap height or
x-height on top, descender or baseline at the bottom. On a real scanned
form, a rewritten "DT" -> "DSI" now matches the scanned word's height
and baseline. The engine OCR-edit tests build their words the way a scan
would box them.
* Edit OCR text: sample colours only for invisible blocks, from the current bytes
Session::apply looked up every edited block with a fresh text_blocks pass and
re-serialized the whole document to render the sampling tile. Text blocks are
now cached per page on the editor (keyed on the unmodified document's source
identity, so any edit, undo or save invalidates them), visible blocks skip
sampling entirely, and the tile renders from the document's current bytes.
Adds a visible stretched-text regression test and documents that editing any
invisible text makes it visible over a painted cover.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* Edit OCR text: name the cached text-blocks tuple (clippy), rustfmt
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Brandon Thomas <echelon@gmail.com>
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* M3: select an image on the page and copy it
Select tool: a click on a raster image selects it (also one drawn through Form artwork, or with
text over it such as an OCR layer); ⌘C copies it as a file in its original format (a JPEG as
stored in the PDF, otherwise a lossless PNG). Right-click ▸ Copy Image copies its pixels at full
resolution, and Copy Image in Original Format the file; Edit text & images has both for the
selected image. Both respect the document's copy permission.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01GWUuTDPKL5wivnSJuCkC8Q
* M3: translate the copy-image strings into Czech, Kazakh, Slovene and Turkish
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01GWUuTDPKL5wivnSJuCkC8Q
* Copy image: keep the clipboard file in a private per-user folder
Copy Image in Original Format wrote into the shared temporary folder
(/tmp/pdfcraft-clipboard) and deleted everything in it first, so another
user on the machine could plant a link or folder there and have PdfCraft
delete or overwrite their target. The file now goes in PdfCraft's
per-user cache folder (or the portable data folder), which is checked to
be a real folder and set to 0700 on Unix (refused when it isn't ours).
Only earlier copies (regular files named '<stem> p<N> img<M>.jpg|png')
are removed, links are never followed, and the new file is created with
create_new so it can't be redirected.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* Tests: escape Windows paths in synthetic GoToR file specifications
canonicalize() returns \\?\C:\... on Windows; unescaped backslashes in a PDF literal string become control characters, so link_open refused the target and main's Windows CI failed. Taken from #1359's helper fix.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude <noreply@anthropic.com>
Co-authored-by: Brandon Thomas <echelon@gmail.com>
* M7: preserve Arabic text order when copying and editing
Record per-line ActualText for shaped added text and preserve its logical order in Edit text. Cover mixed scripts, marks, copy/search, retyping and save/reopen; retain the drawing golden.
Assisted-by: Codex
* Edit text: keep only right-to-left ActualText out of line joining and bidi recovery
Word and Chrome wrap Latin ligatures in /ActualText spans mid-word; splitting
lines at every span broke such words into pieces in Edit Text. Only spans
whose ActualText has strong right-to-left characters keep their own line and
typing order now. Regression: a Latin fi ligature span stays on its line.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Brandon Thomas <echelon@gmail.com>
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* XFA: one script thread per event, and the live form kept up to date in place
A form with hundreds of calculate scripts used to spend its open-time
budget on set-up: every script got a thread of its own, and the engine
dropped its live form after every value a script settled, rebuilding
the whole tree for the next script.
The js crate gains xfa_session::XfaSession: the scripts of one event
run on one thread (with the script stack), each in a fresh JavaScript
context or FormCalc interpreter on a copy of the form as the engine
holds it then, so scripts share nothing, a script that fails halfway
leaves nothing behind, and what a script changed reaches the next one
only through the engine. A script past its time limit is abandoned on
the thread and the session is dead from then on; a fault in the engine
(a panic) loses only that script, the thread is replaced. On wasm the
scripts run inline, as before.
The engine's runner keeps one session per event and, for a value, now
refreshes the one field in its live form in place (xfa::refresh_value,
computing the value as the tree builder would: saved values, template
defaults) instead of rebuilding the tree; rows, presence, access,
resets, relayouts and data rewrites still rebuild it. Under test, the
live form is compared with a fresh build before every script and at
the end of every event, and a difference ends the test binary (a panic
would be caught by the engine's guard and turned into a script error).
210 calculate scripts (one per row of 10 sections x 20 rows, plus a sum
per section), debug build, instrumented inside the engine: the script
event on open 495 ms -> 204 ms (of which live-form rebuilds 217 ms ->
0), on a change 300 ms -> 172 ms; through the tools, layout included,
open 1.27 s -> 0.97 s and a change 0.44 s -> 0.22 s. The rebuild cost
removed grew with form size times script count.
* CI: the GoToR tests' case lists are only grown on Unix, so Windows must not warn on their mut
* M12: say when a form's scripts were turned off
When an XFA script runs past its time limit it is abandoned and the
form's scripts are turned off for the document. The only trace was an
error line in the JavaScript console, so calculations and buttons just
stopped working.
- Engine: Document::xfa_scripts_off() reports the state (set on open and
on an edit), so a viewer shows it without reading error text. doc_info
reports it to agents as xfa_scripts_off.
- App: when it happens (at open or after a click) a toast says so once,
and the notice bar says so from then on, ahead of its other notices,
with a JavaScript Console button. A notice bar the user had closed
comes back for it; closed again, it stays closed.
- The sentence is translated in the 17 catalogs that are complete
(hashes in ATTRIBUTION.toml updated).
Tests: two kittest UI tests (runaway at open; runaway click after the
bar was closed), which fail without the app change; engine assertions
for both states; a doc_info tool test.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* Catalogs: the scripts-off notice in Greek, Polish and Telugu; refresh hashes
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
Co-authored-by: Brandon Thomas <echelon@gmail.com>
* feat(edit): remove all links command and context menu (#539)
- Expose 'edit.remove_links' and 'edit.remove_all_links' commands to remove all links across the document or on specified pages.
- Add right-click context menu options for links: Properties, Delete, and Remove all links in both Link tool and Select tool modes.
- Extend automation tool 'links_remove' with optional 'pages' parameter and removal count.
- Add notification feedback in UI for link deletion.
- Mark 'edit.remove-all-links' as shipped in Parity with evidence tests.
- Add unit and integration tests covering automation, engine, and UI undo/redo.
* Remove all links: one command id, catalog rows, link menu only on a link
- Drop the duplicate edit.remove_all_links command (same label as the
existing edit.remove_links, so it showed twice); edit.remove_links now
runs remove_all_links with its toast.
- Add the toast strings (No links to remove, Removed 1 link, Removed {n}
links) to all 20 catalogs.
- The canvas context menu shows the link items only when the right-click
hit a link (LinkView::context, set per right-click), so a link left
selected no longer takes over the comments menu, and the Link tool no
longer replaces the comments menu away from links. Regression test.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* Restore the PR's links_remove pages test (lost with the workdir conflict)
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Brandon Thomas <echelon@gmail.com>
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* test: give faux-redaction audit its own workdir (parallel-test flake)
* hayro: ColorReplacement type and luminance mapping (reading colours, #897)
* fmt: engine dates (upstream Greek localization miss)
* hayro: apply ColorReplacement to solid text and line-art paints (#897)
* render: ReadingColors view setting flows into RenderSettings (#897)
* engine: per-document reading colours state (#897)
* automation: view_reading_colors tool; page_render honours it (#897)
* ui: Reading colours preference (Replace Document Colors, #897)
* ui: seed reading colours on revision and signed-version opens (#897)
* parity: a11y.replace-page-colors shipped (#897)
* render: map stencil-mask paints through reading colours (#897)
An ImageMask stencil's solid paint was baked into the raster bytes raw,
bypassing the reading-colours hook every other solid fill goes through:
under white-on-black, CCITT scans, JBIG2 text and stencil logos drew
black ink on the black mapped background and vanished. Route it through
replaced(paint, false) — stencils are line art in disguise — and cover
it with an ImageMask fixture test sampling mapped ink on a mapped
background.
* render: print previews render with reading colours off (#897)
Print previews went through the document's render pool, whose config now
carries the reading colours: a white-on-black user saw black sheets in
the Print dialog while the actual print stayed uncoloured. The request
tag is the per-request channel that already crosses the render boundary,
so PRINT_TAG now lives in pdfcraft_render and the renderer honours it by
rendering previews with reading colours off — a preview's job is
fidelity to the printout. The main view stays recoloured. Covered on
both render paths (inline PageRenderer and pool worker).
* ui: seed reading colours on created-document opens (#897)
open_created (comment summaries) and open_created_bytes (images, text,
blank pages, clipboard) pushed new DocViews without seeding the reading
colours, unlike every other open path. A created document now starts
with the app-wide preference too; the seed-on-open test drives both
paths.
* docs: reading-colours flag echo and display-fidelity parity notes (#897)
The view_reading_colors description now says a flag that differs from
the mode's preset default makes the echoed mode read custom, and the
a11y.replace-page-colors parity notes record that print previews and
view snapshots render with the document's true colours: reading colours
apply to the reading view only.
* Merge follow-ups: reading colours beside preview renders and the new catalogs
- Renderer::new carries both color_replacement and the new preview flag
- our raster tests construct RenderRequest with preview: false
- the preference dialog grew past the window: scroll the combo into view
and give the fixture a valid xref so no repair banner covers the dialog
- checkboxes use the crate's square-box widget (clippy disallows egui's)
- Kazakh and Turkish catalogs gain the ten reading-colours strings
* Reading colours: record the hayro patch in vendor/README.md; add missing translations
vendor/README.md gains hayro patch (13): `RenderSettings::color_replacement` /
`ColorReplacement`, the `render_pdf` signature change and the removed
`[dev-dependencies.log]`, with its tests. Adds the ten reading-colours labels
to el, es, pl, te and zh-hant.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Brandon Thomas <echelon@gmail.com>
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* M7: open local GoToR PDF links natively
* GoToR links: read only regular .pdf files under a size cap; add missing translations
A GoToR link in an untrusted PDF could name /dev/zero (unbounded read), a FIFO
(UI thread hangs) or a directory. Both the UI path and the `link_open` tool now
go through `pdfcraft_engine::read_remote_document`, which requires a `.pdf`
extension and a regular file (metadata checked before opening) of at most
1 GiB, and stops reading if the file grows past the cap. Regression tests cover
directories, a /dev/zero symlink and a FIFO (unix) in the engine, the tool and
the UI.
Adds the four link notifications to all 20 catalogs and "The existing action
will be preserved." to el, es, pl, te and zh-hant.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Brandon Thomas <echelon@gmail.com>
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* Gates: fix two clippy failures on main
- ocr_ui: "Also recognize pages that already have text" used egui's own
checkbox, which clippy.toml disallows (#1184 landed beside #1173);
use widgets::checkbox like every other box.
- camera_edges: Rust 1.97's manual_range_contains; the rewrite keeps the
old answer for a NaN angle (false) with an is_finite check.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_019ojKsSCHWk2AuyDTFMHCd7
* M10: print from the web build through the browser's print dialog (gap 4)
The web build had no printer: spool::submit refuses on wasm, so Print
could only download the print-ready PDF. Now its Print dialog offers
"Browser's print dialog" (the default there), which reaches every
printer the operating system knows.
- web_print: PdfCraft draws the print-ready PDF's sheets itself
(engine Exporter), one sheet a frame since wasm has no threads:
300 dpi up to 100 sheets, 200 up to 300, 150 up to 1,000, at most
64 MP a sheet and 1 GiB of PNGs a job. The PNGs become blob URLs
and print from a hidden srcdoc iframe, one sheet per page, margin 0.
- Chromium prints every page at the default @page size, so a sheet of
the paper's size turned sideways is rotated 90 degrees CCW onto it
(as drivers do); other sizes get named @page rules.
- Copies, two-sided and colour are greyed out (the browser's dialog
sets them). If print() is refused (e.g. a sandboxed frame), the
print-ready PDF downloads instead, with the reason.
- pdfcraft-print stays free of web APIs; six new strings in all 20
catalogs (ATTRIBUTION hashes refreshed); docs and parity updated,
gap 4 closed.
Verified in headless Chromium against a trunk build: three sheets at
300 dpi, the landscape sheet turned, no blank trailing page, and the
download fallback. Firefox and Safari not yet tried.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_019ojKsSCHWk2AuyDTFMHCd7
---------
Co-authored-by: Claude <noreply@anthropic.com>
Co-authored-by: Brandon Thomas <echelon@gmail.com>
* Expose tag parent and depth for hierarchy view
* Accessibility: add visual tag properties panel
* Show collapsible tag hierarchy in accessibility panel
* Open checker findings in the matching structure tag
* Translate Tags panel across supported languages
* Tags panel: walk the structure tree once per edit, not every frame
The Tags panel and checker findings that open a tag share a tag list cached per
document and edit generation.
---------
Co-authored-by: Brandon Thomas <echelon@gmail.com>
* Save accessibility review decisions against PDF revision
* Accessibility review sidecars: read them with a 1 MiB cap
The sidecar path is the caller's and may name any file, so reading it
whole was an unbounded allocation. Stop past 1 MiB and refuse an
oversized file as not a sidecar (load fails, save leaves it alone), like
malformed content. Tests cover malformed, binary and oversized files.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Brandon Thomas <echelon@gmail.com>
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* Fix selected comment quick-bar color (#1191)
* M5: skip quick-bar recolour when commenting is forbidden (#1191)
The quick-bar colour swatch now checks allows_annotation() before
restyling the selected comment, matching the canvas and context-menu
gating. Without it, a restricted document showed a "Change comment
properties failed" toast on every swatch pick; the engine already
rejected the edit. Adds a regression test on an encrypted, print-only
fixture that fails without the guard.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01T3DWkcKZcZLbww1oeDbCfb
* M5: quick-bar colour also recolours a selected arrow (#1191)
Arrows are saved as Line annotations with an ending, so tool_for()
reports them as the Line tool and the Arrow tool's swatch never matched
the selected arrow. comments::made_by() treats the Line and Arrow tools
as the same kind. Adds a regression test that fails without the fix.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01T3DWkcKZcZLbww1oeDbCfb
---------
Co-authored-by: Claude <noreply@anthropic.com>
On upright horizontal text of unrotated pages the caret's tip goes to the
nearer character edge on the baseline, sized to the text; elsewhere it stays
at the click. The size is mapped to user space, so /UserUnit pages match.
Parity stays partial: snapping is UI-only.
AI-assisted (Claude Opus 5.5), reviewed by GPT-5.6-terra (Codex CLI).
Native review welcome.
Co-authored-by: Diego Ospina <diegozpy@users.noreply.github.com>
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* M0: keep the camera edge test NaN-safe under clippy's manual_range_contains
Clippy 1.97 flags `d < 0.35 || d > PI - 0.35` in the camera edge hit test. Its suggested
`!(0.35..=PI - 0.35).contains(&d)` would count a NaN direction as a hit, which the two
comparisons did not, so the NaN check is kept explicitly. Behaviour is unchanged.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_011AsuqvnFXhY9ASL1bA2jtm
* M14: edits let go of the previous working file before writing the next (gap 13)
Every edit writes a new working file: the original plus one appended revision. The previous
one stayed alive in the working bytes, the display bytes and the renderer while the next was
written, so after a few edits a 1 GiB file peaked at three copies.
Each refresh path (full, comments/form, content, undo, redo, after save) now releases them
first: the working and display bytes fall back to the original (already held by the editor)
and the renderer is replaced by an empty one until the refresh rebuilds it. A refresh that
still fails after rollback draws the original instead of nothing. Save still writes from the
editor, never from these.
Measured on a synthetic 1 GiB file (release build, Linux): five edits peak at 2.1 GiB instead
of 3.1 GiB, in 8.5 s instead of 16 s. Saving still peaks at three copies; reading from disk on
demand stays blocked on the renderer (hayro parses one contiguous buffer, gap 2). gaps.md,
the parity notes and the roadmap record both.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_011AsuqvnFXhY9ASL1bA2jtm
---------
Co-authored-by: Claude <noreply@anthropic.com>
* #1092: Add editable Organize page-range selection
Generated with Codex.
* #1092: Leave overlapping main CI cleanup to PR #1167
Generated with Codex.
---------
Co-authored-by: Brandon Thomas <echelon@gmail.com>
* feat: presentation (.pptx) export (#1137)
- Implement OpenXML PresentationML (.pptx) generator in crates/export/src/pptx.rs:
- Generate [Content_Types].xml, _rels/.rels, core properties, presentation.xml
- Build slide master, blank slide layout, and DrawingML theme
- Render slides per page: largest text block as title shape, remaining paragraphs as body text box with font size, bold, italic, and color formatting
- Embed images with proper DrawingML picture shapes and relationships
- Generate DrawingML table shapes with column grid and span merging
- Wire OfficeFormat::Pptx into crates/engine/src/compare.rs and export_office
- Register export.pptx command and mark Presentation (.pptx) Ready in catalog
- Support pptx in automation doc_export_office tool
- Wire export.pptx into UI commands and file save dialog
- Add unit and integration tests across crates/export, engine, automation, and ui-egui
- Update create.export-pptx parity feature evidence and docs
* pptx export: saturating slide positions, matching cell-border tags, theme ea/cs fonts, package test
- Heights and the running vertical position use clamped/saturating arithmetic, so a hostile
text size or image rectangle (1e300, NaN, inf) can't overflow i64; the item loop uses get().
- The table cells' a:lnR and a:lnT borders were closed with </a:lnL>, which made every slide
with a table malformed XML; they now close with their own tags.
- The theme's major/minor fonts carry the schema-required a:ea and a:cs children.
- New test walks the produced zip entry by entry (sizes, inflation, central directory), checks
every XML part's tags balance, the required parts, content-type overrides, relationship
targets and that no placeholders are used; a hostile-sizes test covers the overflow.
- Test literals gain main's Block/Image rotation and size fields.
- Parity notes say what is still unverified.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: KrzysztofC <krzysztofc@localhost-live.lan>
Co-authored-by: Brandon Thomas <echelon@gmail.com>
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* M4: rename colliding form field names when pages are imported or combined
Insert Pages from File, Combine Files and the grouped combine registered
incoming form fields under names the destination already used, silently
blurring them into one field (fields of one fully-qualified name are one
field, so their values mix). Importing now walks both sides' fully
qualified names (parent.T.child) — including names that hang only off
page widgets, which /AcroForm never lists — and renames a colliding
field's own /T to the first free variant (Text1-1, Text1-2, …), so its
children keep their place in the tree and the second of two colliding
incoming fields is renamed too.
A field_conflicts policy (rename by default; merge keeps the older
behaviour) and reporting — ImportOutcome::renamed_fields, the combine
outcomes, and a field_name_conflicts(dst, src) pre-scan — let callers
choose and prepare; engine, tool and Combine-tab wiring follow.
* M4: engine and tool wiring for renamed field names
Insert Pages from File carries a field_conflicts policy (rename by
default) and reports what it renamed: a repair-log note names each
old → new pair and warns that scripts referring to the old names were
not rewritten, and the undo step's label counts the renames, so nothing
is slipped in silently. Undo restores the original names, as ever.
The session's combine paths gain reporting variants —
combine_unlocked_reporting, combine_grouped_reporting — returning the
bytes and the renames, and doc_combine takes field_conflicts
("rename" | "merge") and lists renamed_fields [[old, new], …] in its
result when fields were renamed, so agents can find what their scripts
must call the field now.
* M4: the Combine tab warns before combining that field names collide
The Warnings column previews, per file, the form fields whose names
collide with earlier files' and will be renamed (Text1 → Text1-1, …),
computed once per list by a new in-order pre-scan (organize
combine_field_conflicts over the whole list, engine wrapper over the
tab's byte sources; a file that cannot be opened previews as nothing,
as its row already says why). The message is plural-aware and
translated in all sixteen catalogs.
* M4: parity for renamed field names on combine and insert
organize.field-name-conflicts ships: rename by fully-qualified name
(or merge, on request), reported everywhere it happens — repair log
and undo label on insert, doc_combine's result, the Combine tab's
warning column. The notes say what is still missing: the tab's preview
ignores narrowed page ranges, the UI offers no merge choice,
Create ▸ Multiple files renames silently, and scripts and XFA bindings
that reference old names are never rewritten.
* M4: review fixes for the field-name conflict work
Four findings from the PR's code review:
- Clippy: the -D warnings gate failed on the new engine tests (a
format! with no interpolation, two drop(doc) no-ops on a
reference); the branch could not pass its own CI.
- The Combine tab's preview warned that a file split into parts
collides with itself: the preview made one source per row, while
the combine groups parts of one file and imports it once. The
preview now dedupes by group (each group's first entry stands for
the file, as the combine does), and every part shows the file's
preview.
- Collision detection compared /T bytes, so one name written in two
encodings (PDFDocEncoding, UTF-16BE) did not collide — the exact
silent blur this work exists to prevent — and renaming a UTF-16BE
name rewrote it as from_utf8_lossy mojibake. Names are now compared,
suffixed and written as decoded text (PdfString::to_text /
PdfString::text), which is also how the forms crate reads them.
- The engine's combine_unlocked/combine_grouped duplicated their
_reporting twins line for line; they now delegate with the Rename
policy, as organize's own wrappers already did.
* Field-name conflicts: only Combine and Insert from File rename
import_pages (Duplicate, Replace) and extract_pages (Extract, Split) go back to Merge, so
copies of a document's own fields keep their names and a source with two same-named field
roots isn't split apart. Pasting pages into the document they were copied from also merges;
pasting into another document renames as Insert from File does. Regression tests for
duplicate, replace, extract/split (organize), duplicate (engine) and paste (ui-egui); the
PR's rename tests call import_pages_with(Rename) explicitly.
Merge fix-up: main made edit labels a translatable Label, so the rename suffix is now two
wordings in EDIT_LABELS ("Insert pages from {name} — 1 field renamed" / "— fields
renamed"), translated in all 20 catalogs. The Combine warning's @plural row is added to
el, kk, sl and tr. Catalog hashes refreshed.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Brandon Thomas <echelon@gmail.com>
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
Add a Scan tab to Fill & Sign ▸ Add signature. A phone photo of a
signature on paper is framed automatically (all words; table, sheet edge
and dust ignored; a sideways signature turned upright). The frame can be
moved, resized or redrawn, the photo turned in quarter turns, and an Ink
sensitivity slider picks up faint ink. A morphological paper estimate
(van Herk/Gil-Werman closing) removes shadows and uneven light; the
trimmed ink in its own colour on transparency is saved and reused like an
image signature.
- engine: new signature_scan module (SignaturePhoto, PhotoRect, Turn,
ScanError), bounded decoding, no new dependencies.
- Image tab: accepts full-resolution photos (up to 40 MiB, 50 MP) and
scales them down to the saved-signature limits instead of refusing them.
- Photos and images can be dropped on the Image and Scan tabs.
- fill_sign_add: `scan` and `sensitivity` with `path`.
- Translations for all catalogs, parity entry, roadmap log.
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
Co-authored-by: Brandon Thomas <echelon@gmail.com>
* M10: keep invoice overlays out of Word body flow (#773)
Export inferred tables without visible borders. Keep text and images
positioned on pages with artwork, rotated or large text, and retain
each source page's size, crop rotation and units.
Cover invoice conversion through the automation tools, and handle
unavailable Windows symlink setup in the confinement test.
Co-authored-by: Codex <noreply@openai.com>
* M10: add @lollinoo's chosen contributor credit (#773)
The contributor requested their public GitHub profile name while preparing the #773 PR.
Co-authored-by: Codex <noreply@openai.com>
* Word export: float a page only when text overlaps an image or watermark
positioned() turned every page with any image, or with text twice the body
size, into absolutely positioned VML text boxes, so a page with a logo or a
big title lost its flowing paragraphs, headings and tables. Now a page floats
only when it has rotated text, or when an image or a watermark-sized text
block actually overlaps other text (bounded pairwise test). Other pages flow
with images inline. Adds a regression test for a page whose image and title
sit clear of the text, and updates the docs and parity notes.
---------
Co-authored-by: Codex <noreply@openai.com>
Co-authored-by: Brandon Thomas <echelon@gmail.com>
* M0: remove upstream trademark artwork from contributor fork
Use the existing document icon and a plain project label. Remove trademark
asset entries and regenerate attribution as required for modified forks.
Co-authored-by: OpenAI Codex <codex@openai.com>
* M7: group compatible PDF words and hide source text during editing
Join compatible words across text objects and subset-font spaces, recognize
component colours, and reuse subset fonts with numeric space positioning.
Render a read-only background with the selected glyphs removed while the
inline editor is open. Keep one frame and source line breaks, and reset
the layout width when switching paragraphs. Expose the same preview through page_render.
Add synthetic extraction, save/reopen, preview and UI regressions. Record
the remaining mixed-style paragraph and installed-font limitations.
Co-authored-by: OpenAI Codex <codex@openai.com>
* M7: preserve rich PDF regions and preview text edits in place
Group mixed source runs into complete visual regions, retain run styles and source encodings, and use validated embedded source faces in the inline editor. Preserve source baselines and show a single live frame as text grows or shrinks, backed by a read-only draft geometry tool.
Validation: strict workspace Clippy, workspace tests, all 30 WASM crates, asset and parity gates, licensed synthetic edit/save/reopen and desktop checks, and an opt-in four-page private-document regression.
Co-authored-by: OpenAI Codex <codex@openai.com>
* M7: keep cropped artwork from blocking text editing
Track page/path clipping separately from full image placement, use visible selection bounds, and prioritize text hit testing over overlapping artwork. Preserve fixed clips during image transformations and expose visible bounds through the headless tools.
Validation: strict workspace Clippy, workspace tests, 30 WASM crates, asset/parity gates, synthetic engine/tool/UI and native desktop checks, both private-document UI regressions, and all-page local edit/save/reopen/undo on a four-page cropped-artwork PDF.
Co-authored-by: OpenAI Codex <codex@openai.com>
* M7: resize cropped images with their clipping paths
Keep the visible frame and image pixels under the same transform. Preserve the
opposite corner and respond to diagonal, horizontal-only and vertical-only
corner drags without jumping from an off-centre grip. Refuse shared clipping
scopes that would move unrelated content.
Add synthetic engine/tool/UI regressions and opt-in private-file coverage
for 80 corner gestures and 10 grow/shrink save/reopen/undo render checks.
Co-authored-by: OpenAI Codex <codex@openai.com>
* M7: complete source font subsets and preserve clipped editing
Add explicit document fonts, CFF embedding, range formatting and bounded contextual shaping; isolate shared image clips and verify private Chinese edits on all pages.
Co-authored-by: OpenAI Codex <codex@openai.com>
* Revert "M0: remove upstream trademark artwork from contributor fork"
This reverts commit 3aa539b5b69e97419bbab83e6cca3d38835b8de8.
* Test: give the text layer up to 4 s before copying (it flaked on a loaded machine)
---------
Co-authored-by: Kamoliddin Fatkhullaev <komoliddinfatxullayev@MacBook-Pro-Komoliddin.local>
Co-authored-by: OpenAI Codex <codex@openai.com>
Co-authored-by: Brandon Thomas <echelon@gmail.com>
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* Accessibility: edit structure tag properties through tools
* Change tag properties: listed in EDIT_LABELS and translated in every catalog; refresh hashes
---------
Co-authored-by: Brandon Thomas <echelon@gmail.com>
* M5: show actions for selected page text
* Selection actions: literal labels (so catalogs can cover them), no unwrap or unreachable
* Catalogs: Add comment in every interface language; refresh hashes
---------
Co-authored-by: Brandon Thomas <echelon@gmail.com>
* Issue #1028: OCR can read pictures on pages that already have text
Recognize Text skipped every page with text, so a picture of text on a
page that also has real text never became searchable. The engine and
the ocr_recognize tool already had skip_text_pages, but the desktop
dialog never offered it, and turning it off read the page's own text
again and added it a second time, so search found every word twice.
- Engine: always collect the page's own glyph boxes; with
skip_text_pages false, a recognised word at least half covered by
them (an earlier OCR layer included) is left out. OcrPage counts
those words as already_on_page. The default still skips pages with
text.
- Dialog: "Also recognize pages that already have text", kept with the
other choices and used by Recognize text in multiple files too.
- Tools: ocr_recognize reports already_on_page; ocr_recognize_files
accepts skip_text_pages.
- The two new labels in the 15 catalogs that translate the OCR dialog.
- Synthetic tests (a picture of a sentence plus real text) for the
engine, the tool and the dialog; the engine test fails with the
filter turned off. Unit tests for the coverage check, including
NaN, infinite, inverted and overlapping boxes.
- parity, docs/roadmap.md M10 row and the ROADMAP.md progress log.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01JqiZ658tGndQaHao5Q6mdm
* Issue #1028: Czech labels for the OCR text-pages option
The Czech catalog was completed (#1066) while #1028 was in review, and
czech_covers_ui_literals requires every tl! literal: the two new
Recognize Text labels were missing, so the test failed after the merge.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01JqiZ658tGndQaHao5Q6mdm
* Contributors: credit rp6186030 as Raj Patel in the About window
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01JqiZ658tGndQaHao5Q6mdm
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
Co-authored-by: copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com>
Co-authored-by: rp6186030 <25706807+rp6186030@users.noreply.github.com>
Co-authored-by: Brandon Thomas <echelon@gmail.com>
* M14: "Edit with PdfCraft" on the File Explorer menu of PDF files (#682)
The Windows MSI adds a classic shell verb under
HKLM\Software\Classes\SystemFileAssociations\.pdf\shell\PdfCraft.Edit that
runs "pdfcraft.exe" "%1" with the app's icon. It appears whichever app is
the default for .pdf and never changes that default. On Windows 11 it is
under "Show more options": a first-level entry needs an IExplorerCommand
handler and package identity, which an MSI doesn't have.
MultiSelectModel is Document: several selected PDFs (up to 15) start one
launch each, and the existing single-instance hand-off opens them as tabs
of one window.
The verb has its own component, conditioned on a new secure public
property INSTALLPDFCONTEXTMENU (default 1, a welcome-dialog checkbox,
INSTALLPDFCONTEXTMENU=0 for unattended installs), following
INSTALLDESKTOPSHORTCUT. Its menu value is the key path, so uninstall
removes every value. The portable zip still registers nothing.
test-msi.ps1 checks the compiled Registry, Component, Property and
CheckBox rows, and that the .pdf default is never written. Docs: README,
docs/releasing.md, parity notes, ROADMAP log.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* Leave ROADMAP.md to main (its log moved on); the change is described in the PR
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
Co-authored-by: Brandon Thomas <echelon@gmail.com>
skylar taylor-barrick
srt8hellcat2-ui
Márk Bartalos
Raj Patel
Earl Co
Alex
Jeremy Yao
Sir Clangington III.
Marco Arnulfo
ysrdevs
Tanish Manku
Tangu Achilis
Francesco Clemeno
Yousef Mojahid
Bryan Mark
Krzysztof Cieślik
Eduardo Soares
cyfung1031
komo1iddin
Diego Ospina
DWAA1660
Bjorneseth
lollinoo
jcl-bit