The Flatpak/OCI migration left update_flatpak_app and the refusal to replace a
repository-managed Flatpak without tests. A stand-in flatpak on PATH, in child
processes so the environment stays private, covers every scope flag, both
command shapes, a successful update and Flatpak's error text; the guard test
asserts the refusal happens before anything is downloaded.
* 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>
flow_table made a table's first row a header row whenever its look had
header_row set, and the look defaults to Word's (header row on). The look
is an option of the table's style (w:tblLook/@w:firstRow), so a table
with no style, or a style the document doesn't have, got a header row
Word never shows: RTF bolded it and marked it \trhdr (repeat on each
page), HTML made it th, and ODT repeated it. Only Repeat as header row
(RowProps::header) makes such a row a header now; styled tables are
unchanged.
Building a paragraph through insert_text/insert_object at its end walked
every run (and insert_object counted every object before the offset) on
each call, so n appends cost O(n^2): one paragraph of 64k runs took
8.5 minutes. push_text and push_object extend the last run or start a
new one, giving the same paragraph as inserting at len().
push_num_places now writes the digits of round(v * 10^places) directly
when that integer is below 2^52 in magnitude, placing the decimal point
and trimming fractional zeros itself. Below 2^52 the integer is exact
and dividing it by 10^places lands within half a decimal step of the
exact quotient, so the text is the same as the formatter's. NaN,
infinities, scaling overflow and larger rounded values take the
existing formatter path unchanged. Requests for more than nine places
are still clamped to nine first.
Tests compare the output with the formatter path for places 0 to 12
over fixed boundary values (negative zero, ties, 1e15, 2^52, f64::MAX,
non-finite values) and about 144,000 values in all (about 48,000
generated, plus the neighboring bit patterns of each). Widening the
limit to 2^54 makes the comparison fail, for example
-1709448349065955.5 at one place.
On 1,200,000 path coordinates at four places (release build, wall
clock on a Linux host running other builds at the same time, seven
alternating runs), the formatter path took 362 to 373 ms and the
integer path 38 to 44 ms.
The README's glTF "Not read" line was doubled and listed sparse
accessors and tangents, which are read.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A primitive's morph targets are not read, and only one scene becomes the
roots. Both now add a warning: a count of the primitives drawn in their
base shape, and which scene of how many was read.
Closes#1938.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A base64 body with `%2B`, `%2F` or `%3D` escapes was refused as bad
base64, and a plain body lost every escaped byte that was not UTF-8 to
U+FFFD. Escapes are now decoded to bytes first, then base64 when the URI
says so. `uri_decode` is unchanged for file names.
Closes#1936
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Reword decode16 and DdsCompression::None docs, check the 16-bit buffer
length before narrowing for BC1-BC3, and test typeless DXGI codes, padded
16-bit rows, G16R16 masks and decode16 of BC1.
A no-op v17 -> v18 step so FilmCraft can save per-clip camera raw
Source Settings (Interpretation::raw) and older builds refuse such
projects instead of dropping the settings on save. Library part of
storytold/filmcraft#1078 (Part of storytold/filmcraft#1065); the MXF
CinemaDNG part already landed in #1145.
encode16 writes 16-bit pixels uncompressed as R16_UNORM, R16G16_UNORM or
R16G16B16A16_UNORM with the DX10 header, and fits BC4/BC5 endpoints and
indices against the 16-bit values; BC1-BC3 round to 8 bits first.
Texture::decode16 reads these formats (and legacy FourCC 36, wide channel
masks, BC4/BC5 interpolated at 16 bits) without losing precision.
- A bitmap that the item or byte limit refuses is counted in
dropped_images once, however many pieces a clip cuts it into.
- An EMF+ pen whose texture brush was not read is counted in
dropped_images, as a lost texture brush object is.
- An EMF+ clip path or region path whose figures all lie on a line
encloses no area, so it gives an empty area instead of its bounding
rectangle.