* image-codecs: wrap a clipboard DIB as a BMP (#778)
Add bmp_probe::dib_to_bmp (prepend the 14-byte file header, pixel offset
from the header with saturating arithmetic, try_reserve for the output)
and bmp_probe::dib_layout (V4/V5 alpha-mask BI_RGB becomes BI_BITFIELDS,
mask copy after the header). Both are ported from WordCraft's
crates/geom/src/picture.rs so it can drop its local copies. Tests compare
byte-for-byte with the originals over crafted DIBs and every prefix, and
decode wrapped results.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* image-codecs: fix DIB offsets for V2/V3 bitfields and ambiguous mask copies (#778)
- dib_to_bmp no longer adds 12 appended-mask bytes to the pixel offset for 52- and
56-byte BI_BITFIELDS headers, which carry their masks inside the header; the
decoder reads the pixels at bfOffBits, so the offset was 12 bytes too far.
- dib_layout treats 12 bytes after the header that equal the masks as an existing
copy only when more bytes follow than the pixels need, so an exact-length DIB
whose pixels equal the masks gets its copy inserted.
- dib_layout reserves the 12 bytes before changing compression or masks, so a
failed reservation returns the DIB unchanged; the mask bytes are a [u8; 12]
array, not a heap copy.
The differential tests against the WordCraft originals now exclude exactly the
52/56-byte bitfields offset and the exact-length payload cases, and assert the new
behaviour for them.
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
png_probe::pixel_model reads width, height, bit depth, colour type and tRNS
presence from IHDR and the chunk headers before IDAT without inflating iCCP;
PixelModel::bytes_per_pixel and has_alpha give a caller's pixel size and alpha.
webp_probe::has_alpha reads the VP8X alpha flag, the VP8L alpha_is_used bit,
and treats plain VP8 as opaque. Differential tests compare against WordCraft's
png_chunks as an oracle.
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
A conservative pre-decode coefficient budget for hostile JPEG headers: the
larger of JpegInfo::held_coefficients and a whole-picture estimate over every
progressive frame header (SOF2/6/10/14, any precision, sampling clamped to
1..=4), with the costliest frame charged at the first frame's size when the
header can't be followed. held_coefficients is unchanged.
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* Expose supported BMP header recognition
Port the BMP routing check from LightCraft commit 304180a500d1b3f87c8c841f269d305161d6c8b9, preserving minimum length and DIB allowlist.
* Test BMP routing boundaries and unsupported DIB sizes
---------
Co-authored-by: Brandon Thomas <echelon@gmail.com>
camera_raw_mode read a missing SubIFDs offset as 0, so it walked the TIFF
header as the pointer array and could classify a malformed TIFF as RAW.
Skip the entry instead, matching LightCraft's classifier.
* image-codecs: read a raster's resolution from its headers (resolution_probe)
file_resolution(bytes) gives the pixels per inch a JPEG, PNG, WebP, BMP or
classic TIFF records without decoding pixels. JPEG uses the decoder's marker
scan (Photoshop, XMP, EXIF, then JFIF), PNG reads pHYs as the decoder does and
falls back to eXIf, WebP its EXIF chunk, TIFF its IFD0 tags; BMP reads the
info header's pixels per metre. Values outside 1-100000 ppi count as none.
DesignCraft (crates/images/src/resolution.rs) and FigCraft
(crates/format/src/density.rs) each carry their own marker and chunk walkers
for this; they can switch to the shared probe.
Part of #778 and #785.
* image-codecs: resolution_probe reads only a whole nine-byte pHYs, and a zero one leaves eXIf
A pHYs chunk that declares any other length, or is cut off before its
CRC, records nothing, and its data is no longer read from the bytes
after it. A zero pixels-per-metre count records nothing either, as in the
decoder, so the eXIf resolution still answers.
A BMP stores its resolution in biXPelsPerMeter and biYPelsPerMeter
(BITMAPINFOHEADER and later). The image crate writes 0 there and nothing
read the fields, so a 300 ppi image saved as BMP reopened at 72 ppi and an
opened BMP's resolution was lost (storytold/photocraft#13630).
decode now reads the fields into Metadata::dpi, encode writes
Metadata::dpi into them whatever embed_metadata says, and
caps(Format::Bmp).dpi is true. encode_bmp_indexed_with_dpi is the palette
writer with a resolution; encode_bmp_indexed keeps its signature and output.
An Exif signature inside a COM segment was taken for the EXIF APP1, and the copy was checked
only against the end of the file, so a replacement longer than the APP1 overwrote the next
marker. Walk the header segments to the first EXIF APP1 and copy only within its declared length.
On panic = "abort" targets (wasm32) the catch_unwind fallback can't run, so
frames whose first component is sampled below the frame's largest factors
go to craft-jpeg before zune-jpeg is called.
zune-jpeg allocates its whole output before reading a scan, so
decode_jpeg_inks now applies decode's scan-data guard (one byte for every
2048 pixels, counted on scan data, not APP padding) first. The comparison
test's noise is computed in u64, so it doesn't overflow on 32-bit targets.
zune-jpeg 0.5 asserts in its upsampler on some subsampled four-component
frames whose first component isn't the largest. decode now catches that
panic and reads the file with craft-jpeg, as it reads arithmetic-coded and
12-bit files. Refs #778.
A jpeg_inks module for apps that keep a CMYK JPEG's inks the way zune-jpeg
readers gave them: every four-component file taken as Adobe-inverted, YCCK
turned into CMYK as libjpeg does (truncating), only the 8-bit Huffman frames
zune-jpeg decodes (arithmetic-coded and 12-bit files are refused), within
Limits. Also the frame's component count and the assembled ICC profile from
the headers alone. Refs #778.
BC4 is one block of 16 single-channel values and BC5 two of them, the
block BC3 already stores its alpha in, so the alpha block encoder and
decoder now serve all three (BC3's output stays the same).
Reading: ATI1 and BC4U, and DXGI 79 (typeless) and 80, open as gray;
ATI2 and BC5U, and DXGI 82 (typeless, as Pillow writes it) and 83,
open as RGB with blue 0. ATI2 follows Microsoft's mapping, red in the
first block. Signed BC4S/BC5S and DXGI 81/84 are refused, naming SNORM.
Writing: DdsCompression::Bc4 (bc4, ati1) from gray and Bc5 (bc5, ati2)
from RGB, with the DX10 header (BC4_UNORM, BC5_UNORM, one 2D texture),
mip chains included. The encode plan converts other layouts to gray
and RGB, so the fidelity warnings report colour to gray and dropped
alpha; BC5's dropped blue is documented under LossyCompression.
DdsCompression is now #[non_exhaustive] and ALL has six entries.
craft-dds holds the header parsing, the uncompressed and BC1-BC3
readers and writers and the block encoder, with its own pixel layout
and error types and rayon behind a `parallel` feature. It parses the
header first (Texture::parse) so a caller can check the size against
its limits before anything is allocated.
craft-image-codecs reads and writes DDS through it, re-exports its
DdsCompression and DdsOptions and maps its errors onto CodecError with
the same messages, so its API and output stay the same. Apps that only
need DDS can depend on craft-dds without the other codecs and their
licences.
* image-codecs: decode AVIF pixels with craft-video's pure-Rust AV1 decoder
New optional `avif-decode` feature (native targets) reads still AVIF
files: the HEIF/ISOBMFF item graph, grids, clap/irot/imir, alpha
auxiliary images (including premultiplied), 8/10/12-bit 4:0:0, 4:2:0,
4:2:2 and 4:4:4, nclx/ICC colour, Exif and XMP, and PQ/HLG to linear
F32 through cicp::linearize as the HEIF path does.
The container walk follows LightCraft #3017 by @emmertex. Every size is
checked against the caller's Limits before the AV1 decoder runs, the
decoder's memory budget is capped by max_alloc, and the decode is
wrapped in catch_unwind so a decoder panic becomes an error.
Fixtures are tiny ffmpeg/libaom files; the 4:2:0 planes match libdav1d
bit for bit. Adds a decode_avif fuzz target.
Part of storytold/craft-libs#778.
* image-codecs, video: address review on the AVIF decoder
- clap: width and height are unsigned fractions, the offsets have signed
numerators, so 0x80000000/0x40000000 is a width of 2, not -2.
- An item with clap, irot or imir more than once (repeated ipma records)
is refused before any canvas work, instead of rotating 261120 times.
- Matrix coefficients 10 (BT.2020 constant luminance) is refused rather
than converted with the non-constant-luminance equations.
- An alpha plane whose own transforms differ from the colour's is refused
rather than composed in the wrong place; none, or the colour's, still
decode.
- craft-video: Decoder::set_shown_budget lowers the per-chunk byte cap on
shown pictures (never above 1 GiB), and the AVIF decoder sets it to
max_alloc, so a payload of many shown pictures is refused before they
are built.
probe_size returned Unsupported for PNG, JPEG and TIFF, so callers sized
those images with a full decode. Each now reads only its header with the
checks decode runs: PNG's IHDR/fcTL, limits and plausibility; JPEG's
first frame header and EXIF; TIFF's default page directory, camera-raw
refusal and Orientation tag. The size is turned upright for 5-8, as
decode returns it.
In a multipart file the chunks start after the offset tables of all
parts. The deep readers in craft-exr and craft-image-codecs took the end
of the deep part's own table as the start of the chunk area, so an
offset into a later part's table passed the chunk-area check and its
bytes were parsed as a chunk.
Both readers now compute where the deep part's table starts and where
the last table ends with checked arithmetic, and refuse a file whose
tables run past its end before reading any offsets. The earlier tables'
sizes are no longer capped or saturated.
Tests build a two-part file from scratch (a deep part and a 1x1 flat
part, in both orders): the deep part decodes as in a single-part file,
an offset into the second table or one byte before the end of the
tables is refused, and a file that ends inside the second table is
refused.