Support string size debug so it can be used for regular development.
It works be writing values into strings, ensuring the buffer size
given is actually available. Developers can use this with memory
checking tools such as ASAN/valgrind to force an error when
the value used for the size of a buffer is larger than the buffer.
Resolve remaining issue with RNA using BLI_strncpy* in generated
callback functions where the size argument didn't represent the
size of the destination buffer.
This is automatically enabled along with ASAN for the
blender_developer.cmake configuration.
Ref PR !107602.
`sinf(float_angle)` is sometimes producing different results on
x86_64 cpus and apple silicon cpus. Convert to double precision
to increase accuracy and consistency.
Partial fix for #104513. More to come.
This addresses the 'liboverrides still tagged for resync after resync
process' error reports.
Turns out, this situation is (unfortunately) normal in some rare cases,
hwne the hierarchies in the liboverride and the reference data have
drastically diverged.Since the detection of liboverrides to resync, and
which one are valid partial resync roots, happens on the 'old' override
hierarchy, it may not match the actual hierarchy recreated from the
reference data, leading to some part of it being ignored by the resync
code.
There is no real way to avoid this situation, the only thing that can be
done is detect it, and re-process the resync code again when it happens.
From quick limited tests in (extremely dirty/outdated) Pets production files,
most of the time only one resync process is needed. Worst case so far
required 5 reprocessing of the same set of data.
There is no error messages from resync anymore in these tests, so at the
very least resync process should now be significantly more reliable than
before...
This issue tackled here is essentially the same as in ac1ac6be9a
(recursive Purge of unused IDs could delete actually used IDs): handling
of dependency loops is extremely delicate...
This commit essentially rewrites from scratch the code that ensures that
all liboverrides and their dependencies that need it are tagged for resync,
and identifies the partial resync roots.
Dependency loops whitin a hierarchy make this process very complex, as
they can lead to having to recursively process more than once some IDs.
Partial resync roots are also a source of complexity, since some IDs may
be detected a potential roots, and later as dependencies of another
partial resync hierarchy.
This new code generates waaaaaayyyyy less error messages when resyncing
nightmare-like production files (i.e. production files which libraries
have been _heavily_ modified). There are still a few error reports in
some cases though, which are the synptoms of another issue that will be
fixed in the next commit.
While in practice this was probably not a big issue (since usually e.g.
objects used by a nodetree would also be used by other objects or
collections in the liboverride hierarchy), in some corner cases it could
have caused missing 'tag for resync' and improper partial resync roots
detection.
Introduces "Optimal" packing, where the layout is a theoretical
best possible for a given input.
e.g. https://erich-friedman.github.io/packing/squinsqu
Also calls multiple packing algorithms, and chooses the best one.
Fix several issues found while investigating missing browser thumbnails.
TIFF, PSD, and PNG now use their old file check code. This is due to
OIIO not having an early-out check within `.open`. Calling `.open`
required a large portion of the file to be available (more than 8192
bytes). The code here can be removed in the future if/when a new API is
available which alleviates this problem.
PSD files often carry along a large blob of ICCProfile metadata.
Attempting to roundtrip this metadata when we cache the thumbnail to
.png was leading to problems. We suppress this metadata now as ICC
profiles are not supported in general and are quite large.
Lastly, even after the mentioned new API is available, OIIO will want to
validate the full header of some file formats. DPX is the largest at a
full 2048 bytes so we must have this as our minimum now too. OS's should
be servicing this read call just as efficiently as when using 64. I
could spot no performance difference here at least.
This was missed during development because Blender will cache thumbnails
into a special .thumbnails directory and the images I was using to test
had already been cached there.
Pull Request: https://projects.blender.org/blender/blender/pulls/107515
This patch will add support to select all channels between active channel
and last-clicked channel without deselecting previous selection. `Shift` key is
now assigned for the range selection and `ctrl` key to extend the selection. This
changes will make multi-selection similar as outliner tree-elements. New function is
created `animchannel_select_range` to handle the range selection of channels.
Old Differential revision: https://archive.blender.org/developer/D17079
Pull Request: https://projects.blender.org/blender/blender/pulls/104565
The current implementation goes like this
* default to a 0-1 range
* you can set `is_bidirectional` to get a -1/1 range
* if the slider is bidirectional the overshoot goes both ways, otherwise only on the positive side
The issue is that it is not possible to have a 0-1 range but overshoot both ways.
This is useful for example on the "Blend to Default" operator, where 0 means current position and 1 default.
But overshoot also makes sense for this operator since the other way just scales away from default.
So this patch:
* removes `is_bidirectional`
* adds `float range[2]` to `tSlider` and a setter for it
* rewrite some logic to handle arbitrary ranges
* split the `allow_overshoot` flag into one flag for either side
Pull Request: https://projects.blender.org/blender/blender/pulls/107406
This will make further changes for light linking easier, where we want to
build multiple trees specialized for each light linking set.
It's also easier to understand than the stack used previously.
Pull Request: https://projects.blender.org/blender/blender/pulls/107560
* add const where possible
* reduce calls to `ED_slider_factor_get` and instead store that in `const float factor`
* rename variables from camel case to snake case, using full words, to conform to naming convention
* use interpf where possible
no functional changes
Pull Request: https://projects.blender.org/blender/blender/pulls/107610
Conditional blocks introduce scopes that need to be kept track of,
linear code is easier to follow. So use early exiting (`continue`s in
this case) to reduce cognitive load, a number of devs consider this good
practice.
There is no advantage in using BLI_strncpy/BLI_strncpy_utf8 when the
destination string has been allocated and won't be concatenated.
Also no need to calloc memory which is filled by strcpy afterwards.
Previously accessing the extension needed to ensure an extension
on an empty path.
This conflicted with DEBUG_STRSIZE as it was assumed the input was
FILE_MAX, where as it was a small buffer with RNA, see: !107602.
Resolve by separating the function that ensures the extension
with the function that finds valid extensions for a format.
Also pass the size of the filepath to functions that ensure the
extension.
Flickering observed when snapping to edge or face with axis constraint.
It was probably introduced in 4eda60c2d8
The snapping code has a timer, so `t->tsnap.snapElem` isn't always set.
So avoid changing the value of `t->tsnap.snapElem`.
