Don't enable conditionally based on the buildtype.
Further, don't use `add_project_arguments()` to enable the instructions:
this will lead to crashes within g-ir-scanner, which can't properly
parse these instructions.
https://gitlab.gnome.org/GNOME/gimp/-/issues/5053
My previous commit broke the autotools build. Apparently when using
g_object_unref(), some C++ symbol leaked into libapppaint.a archive
library, hence the main binaries (e.g. gimp-2.99) could not be linked
without adding -lstdc++ flag:
> /usr/bin/ld: paint/libapppaint.a(gimppaintcore-loops.o): undefined reference to symbol '__gxx_personality_v0@@CXXABI_1.3'
> /usr/bin/ld: /usr/lib/x86_64-linux-gnu/libstdc++.so.6: error adding symbols: DSO missing from command line
Not exactly sure why using this GLib function in particular caused this,
but let's just try another approach in order not to link the main binary
with C++ standard lib.
Instead let's manage all GeglOperation allocation in gimp-layer-modes.c
by adding a gimp_layer_modes_exit() function and some static array for
storing operation object of each layer mode.
The GimpOperationLayerMode variable member in DoLayerBlend was not
properly constructed. C++ class constructor can be called by creating
object variables, but with GObject, we have to do it with pointers.
Otherwise here we were only allocating the memory for the struct, but
not actually calling any initialization functions.
Also it would seem that the struct was not initialized at zero, as the
space_fish variable was not NULL when it should (i.e. even with same
composite and blend space), hence composite_to_blend_fish was not NULL
and since the operation was not a valid GeglOperation when entering
gimp_operation_layer_mode_real_process(), we crashed.
Not sure how it went unseen for so long!
So instead let's make the layer_mode class member into a pointer. As
such, I have to properly allocate and free it. This is also why I am
adding a copy constructor which will ref the pointer (otherwise we unref
more than we ref as the default copy constructor would just copy the
pointer).
This fixes:
> GEGL-WARNING: (../../src/gegl/gegl/buffer/gegl-tile.c:127):gegl_tile_dup: runtime check failed: (! src->damage)
Which happened when a GEGL operation was running and you canceled it in
the middle, say with the ESC key (easy to reproduce with long operations
such as "Color to Gray"). In such case, obviously don't try to copy the
unfinished operation result into the dest buffer.
... which determines if a node's operation-class has a specific
key, and can be used instead of gimp_gegl_node_get_key() when only
existence is important, to avoid compiler warnings.
Update the rest of the code to use the new function.
... which allows temporarily turning the applicator into a NOP,
without destroying cached data.
This commit also improves gimp_applicator_set_{src,dest}_buffer().
In gimp_tile_handler_validate_buffer_set_extent(), suspend tile
validation while calling gimp_gegl_buffer_set_extent(), so that if
the call triggers clearing of partial tiles, these tiles don't get
unnecessarily validated.
In gimp_tile_handler_validate_validate_tile(), when validating a
partial tile with negative coordinates, make sure to adjust the
result of the modulo when calculating the tile-realtive coordinates
so that they're non-negative, to fix the tile-data pointer offset.
Add a new GimpTileHandlerValidate::invalidated signal, which is
emitted when a region of the buffer is invalidated. This would
allow us to properly invalidate the graph in response; this
normally happens in response to GeglBuffer::changed, but this
signal is not emitted when a region is merely invalidated.
In gimp_gegl_apply_cached_operation(), use gint64 for storing the
total and processed pixel counts used for reporting progress, to
avoid overflowing when applying an operation to a large image.
More of the files were wrong, or at least not absolutely identical to
the files generated by the autotools. I am not doing any code change
other than trying to make both build systems produce identical files
(except for slight differences on 2 files not worth the effort) even
though maybe some things can be improved (especially on the include
list). Maybe to be improved later.
Also fixing 2 of the previously autotools-generated files because of
space typos which should have been committed earlier.
Finally it is to be noted that there is no logics to copy the generated
files back to the source directory in the meson rules. I am not sure
anyway this is really worth it and maybe we should just stop tracking
these generated files eventually.
Restore the behavior of gimp_gegl_apply_operation() prior to
11629fde66 when src_buffer is NULL:
keep the existing operation-node input, instead of using an empty
input. Unlike gimp_gegl_apply_cached_operation(),
gimp_gegl_apply_operation() doesn't have an explicit
connect_src_buffer parameter.
This fixes empty output when merging layers.
In gimp_gegl_apply_cached_operation(), add a boolean
connect_src_buffer parameter, which determines whether to connect
the source buffer to the operation-node's input, or to use its
existing input. In gimp_drawable_merge_filter(), pass FALSE for
connect_src_buffer, so that the existing filter-node input is used.
This produces an equivalent result, however, it avoids invalidating
the filter node, and dropping cached data as a result. In
operations that cache larger areas than the ROI, this avoids
reprocessing already-cached data when processing the rest of the
operation.
Additionally, in gimp_gegl_apply_cached_operation(), use an empty
input for the operation if src_buffer is NULL and
connect_src_buffer is TRUE; previously, we'd use the operation-
node's existing input when src_buffer was NULL. Furthermore, crop
the operation-node's input to the destination rect when crop_input
is TRUE, even if connect_src_buffer is FALSE.
In gimp_gegl_apply_cached_operation(), when applying a non-point
filter with the same source and destination buffers, render the
result to a temporary buffer to avoid chunking artifacts. We'd
previously duplicate the source buffer instead (with commit
35729ee02a erroneously copying the
cached results to the source/destination buffer before duplicating
it, causing this bug), but we now use a temporary result buffer
instead; this has roughly the same overhead, but would allow us to
keep the original operation-node input when committing a drawable
filter in a future commit, which would avoid dropping any cached
data.
Add a new "Swap compression" option to the preferences, allowing
explicit control over the tile-swap compression algorithm.
Previously, control over swap compression was only possible through
GEGL command-line options/environment variables. Since the GEGL
API to list all available compression algorithms is still private
for now, we currently only list the three predefined compression
levels -- "best performance" (the default), "balanced", and "best
compression" -- and a "none" option, to disable compression
altogether. Selecting a custom compression algorithm is possible
by entering its name manually.
In gimp_gegl_buffer_set_extent(), clear the full now-out-of-bounds
region of the buffer, instead of only full out-of-bounds tiles;
however, we still make sure to clear full tiles, instead of partial
tiles, as much as possible. This prevents (parts of) the old
content of the buffer from showing when it's enlarged again. This
is especially relevant for the image projection, once we add
support for a dynamically-expanding canvas in the following
commits, since the projection of a reexpanded buffer can be
temporarily rendered to the display before it's fully
reconstructed, exposing parts of the old content.
.... as a drop-in replacement for gegl_buffer_set_extent(), which,
in addition to setting the buffer's extent, clears any now-out-of-
bounds tiles, to free memory.
In gimp_gegl_apply_cached_operation(), copy the cached results to
the destination buffer *before* reconnecting the operation node, as
this may cause the cache to be cleared (see commit
gegl@4347255cd9d5031e6757c70fdde5c63539d5f508).
which means that it's now included normally via gimpbase.h
and not any longer via gimpbasetypes.h which we only did out
of lazyness. A *lot* of files in libgimp* and app/ now need to
... as a drop-in replacement for gegl_buffer_dup(), which COWs all
tiles of the source buffer, including ones that aren't fully
included within buffer's extent.
In gimp_gegl_buffer_copy(), assume that the area outside the abyss
is empty, and so, when the abyss policy is NONE, allow data otuside
the source/destination abyss to be copied, by temporarily modifying
the abyss. This allows tiles that aren't fully contained within
the abyss to be COWed, unlike gegl_buffer_copy(), which has to be
more general.
In gimp_gegl_apply_[cached_]operation(), use a longer iteration
interval (resulying in bigger chunks) when processing the op, than
the iteration interval used for rendering the projection. In
particular, use an even longer interval when processing area
filters, since their may be particularly sensitive to the chunk
size (see, for example, issue #3711). Likewise, use the asme
longer interval when not showing progress indication, since we
don't need to stay responsive in this case (but don't avoid
chunking altogether, to reduce the space required for intermediate
results).
This allows us to process an op faster when committing a filter,
while still remaining responsive (if overall slower) during
preview.
... which determines if a node is an area filter operation. If the
node is a meta op, we conservatively return TRUE, as it may involve
an area-filter op.
Add a "gboolean edge_lock" parameter to GimpChannel::feather() and a
"Selected areas continue outside the image" toggle to the "Feather
Selection" dialog, just like they exist for shrink selection and
border selection. At the end, convert the boolean to the right abyss
policy for gegl:gaussian-blur.
In gimp_gegl_apply_[cached_]operation(), add support for output
buffers whose extent's top-left corner is not (0, 0). This is
needed by the previous commit.
In gimp_layer_mode_get_format(), disregard the requested composite
space when selecting the format, if the input layer mode is alpha-
only, and the requested composite mode is not UNION, since, in this
case, the layer mode doesn't combine the layer/backdrop colors, and
rather only modifies the alpha of one of them. This allows us to
use the preferred format, avoiding gamma conversion.
This particularly improves the performance of the Eraser tool in
perceptual images.
In the vairous gimp_gegl_create_foo_node() functions, set the
parent node's underlying operation node, so that
gimp_gegl_apply_cached_operation() avoids duplicating the source
buffer when applying these nodes (all underlying operations are
currently point ops.)
In gimp_gegl_apply_cached_operation(), use the underlying
operation, as returned from
gimp_gegl_node_get_underlying_operation(), for testing whether the
operation is a point operation, for the purpose of avoiding
duplicating the input buffer. Likewise, avoid duplicating the
buffer when the underlying operation is a source operation.
... which allow setting/getting the "underlying operation" node of
a graph node. For example, GimpDrawableFilter constructs a complex
graph around a given operation node, which would be the underlying
operation of the graph. This allows querying the properties of the
underlying operation, given only the graph.
In recursive cases, gimp_gegl_node_get_underlying_operation()
returns the most-nested underlying operation; when no underlying
operation has been set, gimp_gegl_node_get_underlying_operation()
returns the input node.
In gimp_drawable_merge_filter(), disable the filter applicator's
cache and output-format conversion nodes before processing the
uncached region of the filter, so that the result is written
directly to the drawable's buffer.
Improve gimp_gegl_mask_combine_ellipse_rect() -- the funciton
responsible for rendering ellipse/rounded-rectangle selections.
Most notably, this commit significantly improves the function's
performance, by identifying whole tiles, whole rows, or parts of a
row, that are fully inside, or fully outside, the ellipse, and
filling them in bulk, instead of calculating the anti-aliasing
value at each pixel, which is now only done along the
circumference.
This commit also improves anti-aliasing, by more accurately
approximating the distance from a pixel to the ellipse, and by
normalizing the distance according to the pixel's cross-section
length in the direction of the said point. In particular, we
guarantee that pixels that are fully inside/outside the ellipse
have a value of 1/0, respectively, facilitating the aforementioned
optimization.
Additionally, this commit fixes various edge cases where several
primitives coincide at a single pixel (in the rounded-rectangle
case), adds support for CHANNEL_OP_INTERSECT, and parallelizes
processing.
Niels De Graef