Prepare reusable soft-mask canvases for filtered and backdrop-dependent masks,
and use a faster destination-in composition path where possible.
Handle Alpha SMask /BC correctly, preserve OOB alpha behavior, and mirror canvas path
operations needed while rendering inside soft-mask mode (mirrored clip was buggy).
Add reftest PDFs covering Alpha masks, transfer functions, backdrop/OOB
alpha, and the optimized composition paths.
This fixes two things that I overlooked in PR 21225, more specifically:
- Use proper, rather than semi, private class fields in `WasmImage`.
- Make tracking of `WasmImage` instances optional, to avoid keeping data alive permanently in the `IMAGE_DECODERS` build.
Given that these classes are, with the exception of their `decode` methods, virtually identical this helps reduce code duplication and simplifies maintenance.
These changes reduce the size of the `gulp mozcentral` build-target by `1292` bytes, which obviously isn't a lot but still cannot hurt.
- Shorten the `getURL` function slightly, by re-factoring the try-catch blocks.
- Change how the `decode` function looks for a decoded ".pdf" name, to skip the regular expression matching when it's not needed and to allow re-using the already defined `pdfRegex`.
This was necessary before charset compilation was implemented, however that's been supported for many years and this is just dead code now.
- PR 9340, back in 2018, stopped using the `raw` field.
- PR 10591, back in 2019, implemented proper charset compilation.
The `CFFFont.prototype.data` should contain a `Uint8Array`, however if compilation failed it was being set to a `Stream` instance which will thus fail elsewhere in the font-code.
*Please note:* This was found by code inspection, since I don't have a PDF document that's fixed by this change.
Return the data as-is from the `CFFCompiler.prototype.compile` method, rather than making a copy of it first.
The reason that it was implemented this way in PR 21053 was to avoid keeping a potentially large `ArrayBuffer` alive, see https://github.com/mozilla/pdf.js/pull/21053#discussion_r3045402988
Having traced all the call-sites in the font-code that directly or indirectly invoke that code, I've now managed to conclude that the compiled CFF-data is never stored on the `Font` instance and using the data as-is thus shouldn't increase permanent memory usage.
This avoids having to "duplicate" dummy `readBlock` methods in a couple of image-stream classes.
Also, move a few `DecodeStream` field definitions to (ever so slightly) shorten the code.
The `textLayerImagePlaceholder` canvas added in #20626 covers scanned
pages and was not recognized as a valid pointerdown target by
`#textLayerPointerDown`, so free highlights couldn't start.
If `PDFDocumentLoadingTask.destroy` ran while `workerIdPromise` was
pending, the inner `.then` in `getDocument` threw "Loading aborted"
before `WorkerTransport` was constructed, so `_transport` was never set
and the "Terminate" message was never posted.
Remove the "Float32Array" mention in the comment, given that the
implementation usesa Float64Array.
Actually using a Float32Array passes all the tests we currently have and
reduces memory usage (by 16 bytes per op), however to be sure that it
does not introduce rounding bugs we'd need to `Math.fround` all
operations we do on the clipBox and pendingBBox. It reduces the
readibilty of the code, but we can revisit if this memory usage becomes
a problem.
When pages carry explicit pageIndices (e.g. after a delete),
resolve insertAfter against that layout instead of the empty
base sequence. Also reject partial pageIndices combined with
insertAfter, which would race against the extraction's auto-fill.
In practice this probably hasn't caused any bugs, however given that `DOMElement.style.zIndex` returns a string the existing code is subtly wrong.
This is also consistent with pre-existing `zIndex` code in the `PopupElement` class, see the `#show` and `#hide` methods.
Currently the hash-property is just a stringified Array, which means that the hash-property can become arbitrarily long. That's not a good idea since it's used to compute a cache-key, in the API, which is then sent to the worker-thread. Hence the hash-property should be reasonably short, and its length should *not* depend on the number of modified editors, which can be achieved by using `MurmurHash3_64` here as well.
Many `parseInt` call-sites already provide the `radix` argument, and this rule helps improve consistency in the code-base; see https://eslint.org/docs/latest/rules/radix
*Please note:* The rule is disabled in `src/scripting_api/util.js` for now, since it's not obvious at a glance (at least to me) what the correct `radix` argument should be there.
The alpha feature is available in Firefox nightly (with the pref `dom.forms.html_color_picker.enabled` set to `true`).
It's available in Safari but not in Chrome.
*Note:* This is similar to PR 19525, which did the same thing for the OpenJPEG decoder.
The advantages of doing this are:
- The same JBig2 decoder is used regardless of WASM being supported or not, which means consistent rendering.
- The old `Jbig2Image` implementation has various bugs and missing features.
- Less code that needs to be maintained in the PDF.js project, since both the CCITT and the JBig2 decoder is replaced.
The disadvantage of doing this is:
- Slightly larger bundle size, however the effect is limited since a fair amount of PDF.js code can be removed. For the `gulp mozcentral` target the size increase is approximately 54 kilo-bytes (which is small compared to the 452 kilo-bytes for the JS version of the OpenJPEG decoder).
