Due to issues with library overrides and unsynced pointers, it's quite
common for the Blender Animation Studio to get crashes of BAT. To avoid
these, Strict Pointer Mode is disabled when using BAT from the CLI.
Blender Cloud add-on will also get a similar update, so that there also
the Strict Pointer Mode is disabled.
Add a 'strict pointer mode' to the `BlendFile` class, which is enabled
by default. This allows users of the `BlendFile` class to decide whether
a bad pointer (i.e. one that points to a non-existing datablock) returns
`None` or raises a `SegmentationFault` exception.
This commit fixes a bunch of issues at the same time, as they are all
related to path handling:
- `pathlib.Path.resolve()` or `.absolute()` are replaced by
`bpathlib.make_absolute()`. The latter does NOT follow symlinks and does
NOT network mounts from a drive letter to UNC notation. This also has
advantages on non-Windows sytems, as it allows BAT-packing a directory
structure with symlinked files (such as a Shaman checkout).
- Better handling of drive letters, and of paths that cross drive
boundaries.
- Better testing of Windows-specific cases when running the tests on
Windows, and of POSIX-specific cases on other platforms.
Thanks to @wisaac for starting this patch in D6676.
Thanks to @jbakker for pointing out the drive letter issue. This fixes
T70655.
Not sure when/how that happens, but it happened on a lighting file of the
Spring project, when linking in a nested collection of which parts where
also linked in from other blend files.
Function declarations that have no type annotations at all will be skipped
by mypy. Adding an explicit `-> None` tells mypy to run on those functions
too.
This does introduce some not-so-nice things, like having to annotate each
`__init__` function with `-> None`. However, the benefits of having static
type checking in a complex bit of software like BAT outweigh the downsides.
Due to the huge number of BlendFileBlock objects created for packing a
production-size blend file, using slots here actually makes the
dependency tracer significantly (p<0.001) faster. In my test case the
speed improvement was 16% for a 'bam list' command.
This simplifies blend file handling, ensuring that blend files are only
opened once. Otherwise it would be harder to handle things like dependency
diamonds (libraries that are referenced via multiple paths through multiple
other libraries).
The expansion process follows pointers and library links to construct
the full set of actually-used data blocks. This set consists of all data
blocks in the initial blend file, and all *actually linked-to* data
blocks in linked blend files.
I've also removed non-recursive dependency listing.
Evaluated only once, so safe to call multiple times without producing
excessive disk I/O. Returns None instead of raising KeyError when there is
no (b'id', b'name') field.
Type refinement is now only done with BlendFileBlock.refine_type(), and
no longer with sdna_index_refine parameters to various functions. This
simplifies the API at the expense of having to call two simple functions
instead of one more complex one.
Sybren A. Stüvel
Sybren A. Stüvel