For example if we're looking for a 2D image named "cubemap", the
importer will tell us there's 0 2D images, making a potential mistake
(2D vs 3D) more obvious.
Using openMemory() instead of openData() allows the implementation to
assume the data will stay in scope for as long as needed, which can
prevent unnecessary copies in some plugin implementations.
It warranted a new flag, DataFlag::ExternallyOwned, to describe this
kind of memory. I couldn't reuse Owned as that's used for allocations
owned by the instance, which is too little for certain future use cases.
For example returning *Data instances referencing an Owned memory would
mean the user has to assume the memory is gone when the importer
instance is gone, and that's generally not true for memory passed to
openMemory().
Originally I thought I would do this later, but then realized the
existing plugin implementations would need to get all updated again to
be aware of the new flag, with some being forgotten, and it's just
easier to do the whole thing in a single step.
This allows to better describe memory ownership and transfer it instead
of forcing the plugins to allocate their own local copy if the import
happens in-place on the imported data. Right now that's mainly for the
openFile() use case, which implicitly allocated an Array with file
contents only to pass it to openData() which then made a copy because it
could not make any assumption about data scope.
In other words, certain plugins (TgaImporter, KtxImporter, DdsImporter,
CgltfImporter and possibly others) will now have their peak memory usage
*halved*.
Hah, so many overloads. Not providing mutable access to keys or layer
offsets as that would break the invariant of the internal array always
being sorted.
Because otherwise we don't properly test all cases. Case in point --
attributeData(UnsignedInt) wasn't correctly propagating the array size,
causing the new tests to fail. Fix in the next commit.
The multi-level APIs still don't check anything regarding level sizes
and order since for example *.ico has no restrictions at all. But the
rest (like non-zero size) is a restriction for all file formats I'm
aware of.
TgaImageConverter test had to be adapted, I expect a lot of breakages in
plugins tests as well. But user code should be fine I think. Also
reduced the rather excessive dimensions in AbstractImageConverterTest,
since the allocation requirement now made the default Emscripten heap
OOM.
To avoid combinatorial explosion of functions that have to be
implemented on the plugin side, the single-image variants proxy to the
multi-level variants, if available.
THe plugin interface string had to be bumped because otherwise there was
a *really nasty* silent ABI break where instead of doSetFlags() it was
suddenly calling doConvert().
The "funniest" of all is that those get printed in a totally random
fashion. Some in Debug, some in Release, some not at all (like, half of
the MaterialDataTest doesn'ŧ initialize `state` and yet it doesn't
complain). Fuck these half-assed features that achieve nothing except
bullying people WHO KNOW WHAT THEY ARE DOING FFS.
Because::This::Is::Colon::Cancer. Also rename Cube to CubeMap and get
rid of the comment that said cube map images are six consecutive 2D
images. Now it's one 3D image instead because that makes more sense, in
the very rare case we'd need to have six different images again we could
probably add a CubeFace value or some such.
First and foremost I need to expand the interface to support 3D
image conversion. But the interface was not great to begin with, so this
takes the opportunity of an API break and does several things:
* The `export*()` names were rather strange and I don't even remember
why I chose that name (maybe because at first I wanted to have an
"exporter" API as a counterpart to importers?)
* In addition, there was no way to convert a compressed image to a
compressed image (or to an uncompressed image) and adding the two
missing variants would be a lot of combinations. So instead the new
convert() returns an ImageData, which can be both, and thus also
allows the converters to produce compressed or uncompressed output
based on some runtime setting, without having to implement two
(four?) separate functions for that and requiring users to know
beforehand what type of an image will be created.
* The ImageConverterFeature enum was named in a really strange way as
well, with ConvertCompressedImage meaning "convert to a compressed
image" while "ConvertCompressedData" instead meant "convert a
compressed image to a data". Utter chaos. It also all implied 2D and
on the other hand had a redundant `Image` in the name, so I went and
remade the whole thing. As mentioned above, two of the enums now mean
the same thing, and are both replaced with Convert2D.
* Finally, similarly as changes elsewhere, I took this opportunity to
get rid of std::string in the convertToFile() APIs.
It should be input first, output second, like with all other APIs. I
remember I was trying something else here, but that didn't really make
sense in the end. Also took that opportunity to get rid of one
std::string.
The original signature is a deprecated alias to the new one and will be
removed in a future release.
Its only use was for specifying N-dimensional SamplerWrapping because,
compared to a Math::Vector, it had an implicit constructor from a single
value (whereas the Vector has it explicit). I solved that by simply
adding a few single-value overloads where it mattered. There, done. No
need for this weird thing and confusion with Containers::Array anymore.
All places that used it now use Math::VectorN<SamplerWrapping>, but the
class is still included for backwards compatibility purposes, together
with providing implicit conversion from and to a Vector.
There will be Flag::FlipY for images at some point, enabled by default
for compatibility with existing GL code, and so it makes sense to start
discouraging setFlags() as early as possible to avoid people resetting
the default by accident.
Also update the imageconverter, sceneconverter and shaderconverter utils
to use these instead of setFlags().
The 5000-line monster took 3.3 second and over 320 MB to compile. While
that may be fine for other projects, that's completely unacceptable here
-- and it seems that this one test is the cause for most of the recent
OOM issues on the CI.
Moreover, I'll be expanding MaterialData with thinfilm, sheen and other
properties that got recently added to glTF and expanding a single test
file with those simply wouldn't scale anymore.
Those were initially implemented and documented when I thought glTF
uses a full cone angle, and I forgot to update them once I discovered
glTF has a half-angle. This is thus now consistent with glTF defaults
again.
Otherwise the attenuation would explode to infinity at distance < 1,
which doesn't make any sense. This, together with the square of the
nominator, is different from the recommended glTF attenuation in
KHR_lights_punctial, but because I couldn't find any relevant discussion
on the equation used there, I'll assume it's just wrong.
Sigh.
Makes more sense as the function isn't expected to fail (and thus any
kind of lazy population is not possible as it would be too late for
error checks anyway).
*Not* updating interface strings even though this is an ABI break
because we're doing that right after the skin import interface bump.
The new materials now commonly import separate per-texture matrices
instead of a single one even if they're all the same because that makes
the plugin implementation *much* simpler. However, existing code that
assumes there's just one matrix would get broken because textureMatrix()
would not return something else. By changing that to return a common
matrix if present and falling back to the global one we can preserve the
original behavior.
Vladimír Vondruš