comp pointer is used for checking if the header is ready.
It can be written in a worker thread and needs a critical
section on the reader side.
This also potentally improve the thread efficiency
from the segment() / marker() function calls.
loader->stack.pop() at line 3271 is called to remove
the defs node added to the stack due to nested graphics elements.
However, it is called in an inappropriate situation and the loader's node tree is damaged.
Fixes an error where the close tag is recognized as a `line` of the graphics tag when it is `linearGradient`.
ex)
```
<defs id="def">
<linearGradient x1="1" y1="2" x2="3" y2="4" id="l"/>
</linearGradient>
<rect x="10" y="10" width="10" height="10" fill="blue" id="r2"/>
</defs>
```
related issue: https://github.com/thorvg/thorvg/issues/2518
Previously, the Lottie builder didn't account for image asset sharing
among multiple layers. This update rectifies the situation.
issue: https://github.com/thorvg/thorvg/issues/2428
Assigning numbers 1 or 2 to a bool always
resulted in true. Now fixed and repeated
shapes are drawn in the proper order.
If the repeater had no elements, iterating
shapes from the end would result in a segf.
This issue is now fixed.
The text node is handled, but default values
of the font-family and font-size as used in
the user's system are not.
For now font has to be loaded by the user.
@Issue: https://github.com/thorvg/thorvg/issues/2350
New api sets/gets the trim of the stroke
along the defined path segment, allowing
control over which part of the stroke is
visible.
@issue: https://github.com/thorvg/thorvg/issues/2190
Layer identifiers do not have to be unique within
the entire file - they are unique within a given
group. Searching the entire composition to find
the referenced matte was an incorrect approach.
Error introduced by 1ee79a6c2a
@Issue: https://github.com/thorvg/thorvg/issues/2349
Shape's property 'direction' should be parsed
together with other properties specific for
the shape. This solves the issue with direction-
it wasn't working.
In the case of a transformation (denoted as T1)
appearing after the repeater, but at the same
level of nesting, the objects to which the repeater
is applied should not consider T1. However,
the transformation of the repeater itself should
take T1 into account.
Also, it is possible to have several consecutive
repeaters, and all of them should be taken into
account. Until now, the last one was overwriting
the previous one.
So far it hasn't been possible to specify
a matte layer - by default, it was the layer
above the calling layer. The 'tp' tag support
has been introduced, enabling referencing any
layer by its index.
In cases where the layer referencing the matte
was the first one, a segmentation fault was
observed. This issue has now been resolved.
@Issue: https://github.com/thorvg/thorvg/issues/2325
Handled the case of different numbers
of points in consecutive frames. This
case can occur due to erroneous data
or as a consequence of changing the path
from closed to open and vice versa.
@Issue: https://github.com/thorvg/thorvg/issues/2287
Rather than aiming for extremely high accuracy,
a compromise can achieve with better performance.
This modification helps prevent unnecessary image rotation.
issue: https://github.com/thorvg/thorvg/issues/2265
- Separate the rounded corner logic from the updateStar
since its logic has become too lengthy and complicated.
- Revise the path generation to eliminate duplicates.
Added support for skew transformation by
a specified angle around a specified axis.
Applied to layers and shapes, not to repeaters.
@Issue: https://github.com/thorvg/thorvg/issues/2262
Implementation of the roundness property
for a polystar. Only the corners that were
not subject to rounding by the inner or
outer roundness parameters are affected.
@Issue: https://github.com/thorvg/thorvg/issues/2230
When resetting back to animated property, system causes an UAF because frames have been freed.
Mark frames in nullptr at the case, so it doesn't use frame data after freed.
Issue: #2255
