* M2: a form XObject that draws itself is drawn once (hayro-interpret patch 89)
A form whose content names itself, directly or through another form, a
pattern or a glyph, drew itself again at every level until the page's
nested-paint budget (50,000) ran out. PDFBOX-5782's poppler-748 page 2,
whose form names a /Pages dictionary as /X0 and so finds itself in the
page's resources, took 38 s.
hayro-interpret now keeps the forms being drawn on the context and skips
a form already on it, reporting ContentTruncated (ISO 32000-2 §8.10.1: a
form is a self-contained description). The lenient fallback from a form's
own resources to the page's stays: MuPDF, Poppler and PDFium all draw a
name from the page's resources when the form's /Resources has no
XObject subdictionary (MuPDF and Poppler also when it has an empty one).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Pq7HyDBXygTf7jcqdsjcfn
* fix(render): a form that reaches itself through a pattern or a glyph is drawn once (hayro-interpret patch 89)
The forms being drawn are now kept per page and thread rather than per
context, so the tiling patterns, Type 3 glyphs, soft masks and groups
painted inside a form, which hayro interprets in contexts of their own,
see that the form is being drawn and skip it inside themselves. A form
drawn again after it has finished draws as before.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Pq7HyCHD1R3rywQBPCHXZA
* chore: re-run CI after the Fill & Sign save comparison failed on the clock (#6066)
`fill_sign_enter_adds_lines_outside_click_commits_and_escape_cancels`
compares two saves whose /ModDate fell in different seconds; it passes
locally on this commit. No change.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Pq7HyCHD1R3rywQBPCHXZA
---------
Co-authored-by: mtlcl410 <chuanlin@niar.org.tw>
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
TIFF predictor 2 (ISO 32000-2 §7.4.4.4, Table 10) went through the
bit-level BitChunks/BitWriter path for every bit depth, re-reading the
previous output pixel through a new BitReader per pixel. Whole-byte
samples (8 and 16 bits per component) are now undone on byte slices,
each sample added to the same component of the pixel to its left,
wrapping at its width; 1-, 2- and 4-bit samples keep the bit-level path.
The output is byte-identical, including the dropped partial last row.
A page of 45 full-page 8-bit grey images so predicted (PDFBOX-1350)
renders in about 4 s instead of 23 s with the same pixels.
Vendored hayro-syntax patch 42.
Claude-Session: https://claude.ai/code/session_01Pq7HyDBXygTf7jcqdsjcfn
Co-authored-by: mtlcl410 <chuanlin@niar.org.tw>
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
Tests only: a JPEG 2000 failure kept for one size is still kept after drawing at another, and
the second page draws the picture decoded on the first and the one the first page's allowance
left out, through one cache.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Pq7HyCHD1R3rywQBPCHXZA
Tests only: the search for the smallest JBIG2 allowance that draws the first picture renders the
first page alone and halves the range, about ten draws instead of up to two thousand.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Pq7HyCHD1R3rywQBPCHXZA
The failure is kept per key, and a JPX decode's key includes the size it
decodes for, so the same failed picture drawn at two sizes is decoded once
at each. The test fails if the cache key leaves the target out.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Pq7HyCHD1R3rywQBPCHXZA
The fallback to the glyph procedures' d0/d1 widths keeps such a page
readable, but ISO 32000-2 Table 110 requires /Widths in a Type 3 font,
so the reader should also be told the font is malformed. Selecting a
Type 3 font without /Widths (or /FirstChar, /LastChar) now reports
InterpreterWarning::Type3WidthsMissing, surfaced once per page as
RenderWarning::Type3WidthsMissing: `pdfcraft-cli render` prints it and the
automation render tools list it with the page's other warnings.
Test: raster::tests::type3_proc_widths::a_type3_font_without_widths_is_reported
(fails with the report switched off).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Pq7HyCHD1R3rywQBPCHXZA
The decoded-image cache stored a failed decode as an empty entry and read
an empty entry as a miss, so a picture or stencil mask that doesn't decode
(damaged data, a JPX or JBIG2 stream its decoder refuses) was decoded again
at every draw, and at every tile, preview and zoom of a render worker that
shares the cache.
The cache now keeps a failure for its key, with the warnings the decode
reported; a later draw reports the same warnings and gets nothing, as the
decode did. A failure counts 128 bytes towards the cache's 64 MiB budget.
A failure the page's JBIG2 allowance may have caused (ContentTruncated, or
the allowance spent) isn't kept, since the next page starts afresh.
Successful decodes are cached as before.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Pq7HyDBXygTf7jcqdsjcfn
decoded_images_serve_every_size_they_are_drawn_at asserts image and mask
decodes separately, through a new InterpreterCache::decoded_mask_decodes
(hayro-interpret patch 82), so a wrong split behind the same total fails.
The truncated-stream case of
an_averaged_down_soft_mask_follows_its_samples_and_its_drawn_size asserts
the exact mask handed over: 32 x 10 (the decoded rows), all opaque.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Pq7HyDBXygTf7jcqdsjcfn
A Type 3 font without /Widths (PDFBOX-2943 misspells the key) gave its
glyphs no advance, so the text was drawn as one blot and extracted as
characters stacked on one spot. ISO 32000-2 Table 110 requires /Widths,
and 9.6.4 has the glyph procedure state the same width with d0/d1, so
such a font now takes that width, read once per code. A font with
/Widths is unchanged; a code outside FirstChar..LastChar keeps width 0
as Table 110 says. Rendering and text extraction share the advance.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Pq7HyDBXygTf7jcqdsjcfn
A Times New Roman the PDF doesn't embed, written as a Type 0 font with a
CIDFontType2 descendant (TimesNewRoman,Bold, TimesNewRomanPSMT and the
other faces) and no /ToUnicode, now goes through the rule of patch (76):
each CID gives, through /CIDToGIDMap (ISO 32000-2 Table 115), a glyph
index in the author's font, and the substitute draws that glyph by its
name in the standard Macintosh order of the TrueType post table. Before,
the CID was taken as one of the substitute's own glyph indexes, which
ISO 32000-2 9.7.4.2 notes "are not defined consistently from one font to
another", and the ameli.fr mutation form drew "Salarié(e) du régime
général" as "Salari²*g+f w't²i lo g'ì²p²tcn".
Times New Roman numbers its first 258 glyphs in that order: the same
form's embedded Times New Roman subsets keep their glyph indexes, and 941
of their 942 /ToUnicode entries below 258 name that order's character.
Times and Times New Roman MT Std are other fonts and keep the old path.
Fixes#6449
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Pq7HyDBXygTf7jcqdsjcfn
hayro patch (30) gains a test hook, `offscreen_draws`: per thread, the tiling cells drawn
and the pixels of every pixmap drawn offscreen, all made by `offscreen_pixmap`. The two
wall-clock bounds become counts: 5,000 patterned stencils draw one cell and less than a page
of pixels offscreen, and 3,000 fills of one setting draw one cell. New tests interleave the
settings of two coloured patterns with cells of one size and one of them at two scales, each
showing its own pattern in place and drawing one cell per setting, and paint an uncoloured
pattern in two colours and, at half opacity, both as a fill and as a stencil mask's copy.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Pq7HyDBXygTf7jcqdsjcfn
CodeRabbit on #6409:
- a kept procedure holds its bytes within the page's content budget while the
glyph is painted, as a decoded one does (context::hold_content), so a page
skips the same glyphs once the budget is spent, kept or not;
- one decoded with no more than 64 KiB of budget left, which the budget may
have cut short, isn't kept;
- each kept procedure counts its name and 64 bytes besides its bytes against
the cache's 32 MiB.
Tests for the 64 KiB and 32 MiB limits and for two fonts with the same glyph
name.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Pq7HyDBXygTf7jcqdsjcfn
A Type 3 font now keeps each glyph procedure it decoded, so a glyph drawn many
times isn't inflated again for every copy. A real file uses a one-glyph Type 3
font as a fill pattern and draws that glyph about 650,000 times on a page; it
rendered in 11.2 s and now in 5.9 s, pixel for pixel the same. A procedure over
64 KiB isn't kept, and the fonts of one InterpreterCache keep at most 32 MiB
together, as fonts live as long as the cache.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Pq7HyDBXygTf7jcqdsjcfn
A font whose head.indexToLocFormat holds neither 0 nor 1 (256, the bytes
of 1 swapped) had its long loca offsets read as short ones, and every
glyph came out wrong: "LINE UP" drew as boxes. The format is now taken
from the loca table's length.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Pq7HyDBXygTf7jcqdsjcfn
`has_larger_mask` read only /Width and /Height, while `ImageXObject::new`
also takes the abbreviations /W and /H, so a large mask sized that way
cached its image's raster without the drawn size and a later, larger draw
reused the small mask.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Pq7HyDBXygTf7jcqdsjcfn
Review follow-ups for hayro patch (30):
- The stencil coverage that cuts the pattern is drawn with in_type3_glyph
cleared, as the sub-render it replaced was: inside a glyph a shrunk
stencil was nearest-sampled and lost strokes thinner than the shrink.
- The bordered cell of a pattern whose step shows one cell (patch (20))
goes through the same cache (Renderer::tiling_cell, keyed apart by
one_cell), so its fills no longer draw and keep a cell each.
- A test pins a patterned stencil under a soft mask to the soft mask's
opacity once.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Pq7HyDBXygTf7jcqdsjcfn
The raster cache key now includes the drawn size, so a soft mask drawn at
several sizes was decoded in full once per size: issue16263's rotated and
scaled copies hit four sizes and took about 10 s. Sizes are rounded up to a
power of two, and every averaged-down size is made from one cached
intermediate mask at most 4096 on a side, so the full mask is decoded once.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Pq7HyDBXygTf7jcqdsjcfn
A stencil mask (/ImageMask) painted with a pattern was drawn through a
luminance mask, which vello_cpu only takes at canvas size: every draw
rendered, converted and composited a full-page mask. And a tiling
pattern's cell pixmap was drawn again for every fill, each copy kept
until the page was composited. pdf.js issue 8078 paints 73,047 16 x 16
stencils with one pattern whose cell is 3.5 MB at 36 dpi: the render
ran for 5 minutes and then failed to allocate.
hayro patch (30): the stencil is drawn under a clip of its device
rectangle (padded by two pixels for a stencil snapped to whole pixels),
the pattern in an isolated layer cut to the stencil by a layer holding
it composited with Compose::DestIn (ISO 32000-2 8.9.6.2: the pattern
shows where the mask's samples paint). A coloured tiling pattern's cell
is drawn once per setting of the pattern and cell scale (TilingCells,
eight entries per renderer). hayro-interpret patch (78) numbers each
setting of a tiling pattern (TilingPattern::instance).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Pq7HyDBXygTf7jcqdsjcfn
is_arial compared the family and style without regard to case but
stripped the MT suffix case-sensitively, so arialmt and Arial-Boldmt
weren't read in Arial's glyph order (CodeRabbit review on #6210). The
suffix is now stripped in any case, slicing only at a char boundary.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Word and PDFMaker write an Arial they don't embed as a Type 0 CIDFontType2
with Identity-H, often without /ToUnicode, so each code is a glyph index in
the author's Arial. hayro drew such a font with the standard CFF substitute
and, with no Unicode route, took each code as one of the substitute's own
glyph indexes: "Low Power" drew as "@ck 'Dck Yf".
Arial numbers its first 258 glyphs in the standard Macintosh order of the
TrueType post table (L is 47, o 82, w 90). For a non-embedded CIDFontType2
whose descendant /BaseFont is one of Arial's four faces, and whose CIDs are
not of an Adobe character collection, the CID is now taken through its
/CIDToGIDMap to that order's glyph name, and the substitute draws its glyph
of that name (.notdef when it has none). Arial Narrow, Black, Unicode MS
and Rounded, and CIDFontType0 fonts, keep the old path. The names come from
read-fonts' post::DEFAULT_GLYPH_NAMES, already in the tree through skrifa.
Vendored hayro-interpret patch 76.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* M1: a crypt filter selected by reference is used
A stream can name its own crypt filter, and every part of that can be indirect. cos read only the
direct forms, and took anything else for no name, which is /Identity. So a stream selecting a filter
by reference was saved in the clear in a file meant to be encrypted, and in a file from elsewhere
its ciphertext was read as data. The /Encrypt dictionary's /StmF, /StrF, /EFF, /CFM, /Filter,
/EncryptMetadata and a filter's /Length had the same fault, and /StmF by reference left every
stream undecrypted. A stream's /Type and a signature dictionary's are read through references too.
transform now reads these from the document, so a full save encrypts before it renumbers: the
references in what it reads have to be the document's. Dropping a crypt filter that new protection
doesn't define comes before the walk, so what it copies out of a reference is encrypted too.
Loading can re-enter here, and LoadGuard ends a cycle, which reads as absent. A chain longer than
any decoder takes names no filter: its entries would be followed for every stream sharing it.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01AmsdeUCdTaD72jkDw2pJKY
* M2: a stream with a crypt filter of its own is drawn
The renderer decrypted every stream with the document's filter and then failed on /Crypt as a filter
it couldn't decode, so a stream naming its own filter read as nothing, however it was spelled, and a
page whose content was such a stream stayed blank. It now uses the filter the stream names, /Identity
without one, and /Crypt is nothing further to decode (hayro-syntax patch 30).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01AmsdeUCdTaD72jkDw2pJKY
* M1: tests for a crypt filter's place among the others
A lone parameter dictionary is the first filter's, so /Crypt behind another filter has no name of
its own, and an array too short for /Crypt's entry stays as it is when the filter is dropped.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01AmsdeUCdTaD72jkDw2pJKY
* M1: encryption reads every legal form, and a full save keeps them
From a review of the crypt filter fix.
The renderer read a stream's own filter from the form /DecodeParms had to match the form of /Filter,
so [/Crypt] with a lone dictionary, an everyday spelling, read as /Identity and the page stayed
blank. Parameters are now taken in either form, which also gives [/FlateDecode] its predictor.
It also refused a V4 file without /StmF or /StrF, which are optional, and used a /CF entry named
/Identity, which is never one. A chain past what cos decodes names no filter there, so it names none
here either (hayro-syntax patch 30).
A full save lost what a direct /Encrypt refers to, and kept its references under the old numbers, so
nothing in the file read afterwards. /O or /U given by reference were decrypted and encrypted again
under a new number, so the file it wrote no longer opened. What /Encrypt and /ID refer to now stays
as stored. /ID by reference, which has to be direct, is read, and written direct.
A signature dictionary is known by a /ByteRange given by reference, and a null or dangling /Type or
/Subtype is an absent one. References to references are followed in Document::resolve, so /Name by
way of two no longer saves a stream in the clear. New protection drops every /Crypt in a chain.
An object stream's /DecodeParms can be given by reference.
Reading what decides how an object is treated can load more objects, each doing the same, which a
file could make a tree that doubles with every level once the cache lets go of it. What is loaded
for one of these no longer follows its own.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01AmsdeUCdTaD72jkDw2pJKY
* M1: what a read follows is bounded by a budget, and nothing shared is copied per stream
From a review of the fixes before this, all in them.
A flag that kept what is loaded for a selector from following its own stopped the fan-out, but a
legal file lost object stream members to it: read first through a selector, an object stream with
parameters by reference was kept undecoded, and a full save wrote nulls without saying so. A budget
replaces it: one read or write of an object, with all it loads on the way, follows up to 32
references, which is more than any object needs when all of it is by reference. Past that they
read as absent, so a file's ladder of selectors still stops, and nothing a real file holds is lost.
Finding out whether a dictionary is a signature's loaded objects for every dictionary of an object,
and a load that fails takes a scan of the file, so one read could hang. Only a dictionary with a
direct string /Contents is asked now, and the budget bounds those.
/ID written direct copied every entry, so a 1 MB file naming a large string thousands of times ran
out of memory: it is at most two strings now. Dropping a crypt filter copied a shared /DecodeParms or
/Filter into every stream, 1 GB from a 1 MB file: of what's shared only filter names and the
parameters decoders read are copied now, through Document::with_direct_parms.
The renderer skips /Crypt in a chain instead of copying the data for each entry.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01AmsdeUCdTaD72jkDw2pJKY
* M1: only an object itself follows references, and a failed scan is remembered
From a review of the budget.
Following a reference while walking deep inside an object added that depth again for each object
loaded, which overflowed a 2 MiB stack on a file under 10 KB and took the process with it, past any
catch_unwind. References are now followed for the object itself and never for what is nested in it.
A stream can't be nested, and a signature dictionary nested in another object reads as it says
directly, as it did before this PR. walk is back to following nothing.
A reference that fails to load scans the file, and nothing remembered that, so a 1 MB file took
minutes to save. What a scan didn't find is kept for the document: the file doesn't change.
Filter names copied out of a shared chain are bounded, since a name can be any length. /ID keeps each
entry in its place, or the key made from the first entry changes. direct_id's comment was on another
function.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01AmsdeUCdTaD72jkDw2pJKY
* M1: tests with no deadline a busy machine can reach
Wide margins on the two that time what they guard. A test of dictionaries nested in an object goes:
nothing nested follows a reference now, which the small-stack test covers.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01AmsdeUCdTaD72jkDw2pJKY
* M1: a crypt filter past what a read may follow is the document's own
Past the budget a reference reads as absent, and a crypt filter with no /Name is /Identity, so a stream
that selected one by reference had its ciphertext read as data. What can't be told takes the document's
filter, as a /Filter that can't be read does.
Found by CodeRabbit.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01AmsdeUCdTaD72jkDw2pJKY
* M1: a read refused, not a budget full, is what leaves a crypt filter unknown
My last commit took a read that used up all it may follow for one that was refused, so an /Identity
stream read at exactly 32 follows was decrypted to garbage. Refused reads are counted now, and only
one on the way to this stream's /Name says anything about it.
Found by CodeRabbit.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01AmsdeUCdTaD72jkDw2pJKY
* M1: a save keeps the handler, the key and the permissions
Under every algorithm and by either kind of save, `/Encrypt`, `/ID`, the file key and the permissions
come back as they were. No test held that. `/ID` by reference is held through an incremental save
too, which leaves it as it was.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01AmsdeUCdTaD72jkDw2pJKY
* cos: encrypt an edited or added object whatever /Encrypt names
stored_as_is now also requires the object to be unedited, so an edit of
an object /Encrypt or /ID refers to, or a new object given a number they
name, is encrypted on both saves. encryption_members skips numbers with
no object behind them and streams, which those dictionaries never hold.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
Co-authored-by: FredApps <153173854+FredApps@users.noreply.github.com>
Review of #3858 (aiongg):
- An area-shrunk image with a soft mask, a stencil's ink included,
built its full-resolution RGBA before shrinking. A 10718 x 13701
stencil at page size peaked at 1.16 GB (587 MB of RGBA). It is now
premultiplied a source row at a time as it is averaged, and the
shrink works a target row at a time, so it peaks at 1.03 GB, as the
nearest-neighbour path did. Same bytes; about 5% slower per shrink.
- An image whose f32 scale passed 1 only by rounding was smoothed
bilinearly and blurred by half a pixel at a half-pixel offset. It is
now smoothed only above 1.01.
- A shrunk /ImageMask stencil keeping its thin strokes (#4800) is now
tested.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A soft mask no larger than its area-shrunk image is sampled onto the
image's grid, and it was always sampled nearest-neighbour: a mask that
asks for /Interpolate became a hard step where main drew a ramp
(jpeskov's 64 x 64 image at 32 x 32 under a 2 x 2 mask). An interpolated
mask is now blended bilinearly onto the grid, centres aligned; one
without /Interpolate is still sampled nearest, which keeps the
colour-fringe fix.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Scanned pages looked thin and broken at screen size. An image that does not
ask for /Interpolate (most scans) was pre-shrunk with nearest-neighbour since
about one sample in ten and dropped most strokes of its text, and the nearest
sampler then read the shrunk raster again at a fractional offset.
- Shrinking (vendored hayro patch 21): non-interpolated images outside Type 3
glyphs are area-averaged (area_shrink: each device pixel the
coverage-weighted mean of the samples it covers, 16.16 fixed point whose
weights sum to exactly one, premultiplied when masked) to their snapped
device rectangle and drawn 1:1. /Interpolate governs smoothing when an image
is enlarged (ISO 32000-2 8.9.5.3); averaging adds nothing between samples
and drops none. PicScale's Box/Area filters were not usable: shrinking, they
average a window four target pixels wide.
- Enlarging (patch 28): an image of page-scan size (512+ samples on its
shorter side) is drawn bilinearly when enlarged, so a zoomed scan has no
staircase edges; icons, QR codes and pixel art keep their hard pixels.
- A snapped image whose shrink failed keeps the full transform instead of
being drawn at its source size (latent in the interpolated path too).
Measured on a 13-page 300 dpi scan at 96, 120 and 192 dpi: every rendered
sample is within 1 of the exact area average (it was up to 255 off, 17-23%
of samples by more than 1). Bilinear zoom costs nothing measurable (four
1024 px tiles at 400%: 50 ms, nearest 48 ms; bicubic was 267 ms).
Review fixes:
- a tile snaps a shrunk image's half-pixel edges where the whole page does
- a scan shrunk one way is smoothed the way it is stretched
- a smaller soft mask doesn't let transparent colour through the area shrink
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Windows exporters map the hyphen of an Arial they don't embed to U+00AD in
/ToUnicode. The CFF standard substitute has no soft hyphen, so the CID was
taken as its glyph index and drew an unrelated glyph (an Ë, or ! in bold).
ISO 32000-2 Annex D.2 calls the soft hyphen typographically the same as
the hyphen, as patch (30) already reads WinAnsi 0xAD: draw the
substitute's hyphen. The text keeps U+00AD.
* A rebuilt xref finds an object whose header names another generation (#5580)
`poppler-85140-0.pdf` is 340 bytes with no cross-reference table, and the
header of its one page reads `3 18446744073709551616 obj` while the page tree
points at `3 0 R`. Before #427 the digits wrapped and 2^64 wrapped to exactly
0, so the header read as `3 0 obj` and the file opened by accident. Once an
overflowing integer became a real, the object was unreachable and the file
stopped opening at all, which `cargo xtask check` reports as a regression.
The generation is not really the point. `3 99999 obj` with the same page tree
is refused in exactly the same way on main -- the control for the new test
fails on that row first, with "page 1 does not exist" -- so this is a
mismatched generation in a rebuilt table, not an integer overflow. A table
rebuilt by scanning the file has no revision history to tell two generations of
an object apart, and a file that needs rebuilding references everything as
`N 0 R` whatever its headers say.
So, in the reconstruction path only:
- a generation number that cannot be read is taken as 0, the value such a
file's references use (`read_object_header`). A generation is at most
65,535 (ISO 32000-2 §7.5.4), so there is nothing to lose by it;
- every scanned object is also filed under generation 0, with `or_insert` so
a real `N 0 obj` is never displaced by the alias;
- the object is read back behind the same lenient header, since reading it
parses the header again and the strict identifier check rejected it
(`read_repaired_body`, gated on `repaired`).
Reading numbers anywhere else is unchanged, and a table read from the file's
own xref keeps the exact identifier match. The scan already files objects found
in object streams under generation 0, and cos's table parser already takes an
unreadable generation as 0, so this follows what the code does elsewhere.
The test covers generations 0, 99999 and 2^64 with no table and no `startxref`,
and asserts the page is found at its own size. Its fixture is built in the test
rather than taken from the corpus, so it holds without the pinned checkout.
* The generation-0 stand-in is one key, not a second object
Two consequences of filing a scanned object under generation 0 as well,
both from review.
`XRef::objects` walks the table's keys, so an object with a stand-in came out
once per key: a document whose page tree cannot be read is recovered by looking
for objects that are pages, and it reported the page twice. Enumeration now
yields one object per location, after the sort it already does. Both keys stay
for lookups; only this walk is deduplicated.
And `or_insert` never replaces, so a later `3 1 obj` moved the exact `3 1` key
to the later body while the `3 0` stand-in stayed on the earlier one -- `3 0 R`
and `3 1 R` would then read different objects. The scan now tracks which
generation-0 keys are stand-ins: a real `N 0 obj` takes the key and keeps it,
and a later header for the same number replaces a stand-in.
The tests check both through what comes out: page 2 must not exist for a
single-page file recovered without a page tree, and with two `3 1 obj` headers
the page comes back at the later body's 300x150 rather than the earlier
200x100.
* M2: the repair scan keeps its place on a non-header, and an object stream's first member is enumerated (#5580)
- read_object_header's lenient path works on a clone, so `42 3 0 obj` no longer costs the
header after the stray number (CodeRabbit, minor).
- XRef::objects deduplicates by location and object number: an object stream and its first
member share (offset, 0) and are two objects (CodeRabbit, minor).
- vendor/README.md: hayro-syntax patch (34) and its tests.
Both new tests fail with their fix reverted.
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>
* M2: a shrunk image on a turned or mirrored page snaps to whole pixels (#5530)
hayro patch 14 (`snap_minified_image`) shrinks an interpolated image once
to its whole-pixel device rectangle and draws that 1:1, instead of
shrinking to a rounded-up size and resampling again at a fractional
scale and offset. It only accepted an upright transform (a > 0, d > 0,
b = c = 0), so every image on a page under /Rotate 90, 180 or 270 -- and
any image placed turned or mirrored by its own matrix -- still took the
double resample. That is #5530: a scanned page that matches PDFium
upright drifts and softens when the same page is rotated.
The snap now accepts every transform whose axes lie along the device's
(the eight quarter-turns and mirrors) and returns the 1:1 transform
that draws the shrunk raster from the snapped corner, one source-axis
step a whole device pixel, in whichever direction that axis runs.
Skewed and freely rotated images keep the old path.
On a synthetic page (crop box and scale on whole pixels, image at
fractional coordinates), /Rotate 90, 180 and 270 renders turned back
differed from the upright render in 72% of pixels, by up to 200 grey
levels; now they are identical. Non-interpolated images were already
identical and stay so.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* M2: snap only when the dropped skew is invisible, and only a resize that succeeded
Review of #5640:
- The snap judged an image axis-aligned by a tolerance on its
coefficients (1e-4 of the scale). For a very wide image a skew that
small still moves the far edge by more than a pixel (40 000 samples
at 4e-5: 1.6 px), and the snap flattened it. It now measures what it
drops: how far each off-axis term moves the far edges, at most a
hundredth of a device pixel (SNAP_SKEW). Upright and turned alike.
- A resize that fails hands back the source raster at source size; the
1:1 snapped transform then drew it at that size. The snapped
transform is now used only when the shrink produced the snapped
size; otherwise the full transform, as before.
- The mirrored case of the edge test drew a uniform image, which looks
the same mirrored; it now also mirrors a half-black image and checks
which side the black lands on.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* M2: the turned-page test draws every kind of image the snap reaches
The /Rotate oracle test drew one 8-bit grey image. Stencil masks (CCITT
scans often are one), 1-bit images and images with a soft mask reach
the snap through other paths, so it now draws all five: 8-bit and
1-bit grey, a stencil mask, a soft mask of the image's size and one of
another size. On main, every interpolated kind is off by 113-217 levels
over 63-73% of the image when turned, and so is a non-interpolated
image with a smaller soft mask (110 levels, 16%): hayro patch 15
filters such a mask whatever /Interpolate says, so it is snapped
upright and was not when turned. All are identical with the fix; the
other non-interpolated kinds were already identical and stay as guards.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
Co-authored-by: Brandon Thomas <echelon@gmail.com>