Blender Git Loki

WM: fallback to regular writing if auto-save can't access undo data

While this is very unlikely, always write the autosave file,
even if the `memfile` undo data is unexpectedly NULL.

Also use CLG for logging warnings.

Cleanup: update comments, remove ifdef'd code

Cleanup: rename G_FILE_RECOVER -> G_FILE_RECOVER_READ

No functional changes, make room for a write equivalent.

Writefile: only include recovery info in auto-save & quit.blend

Previously all blend files included the path they were saved,
causing files distributed publicly to include users local paths.
This also included developers home directories for startup & userprefs
defaults & app-templates bundled with Blender.

Now recovery information is only written for auto-save & quit.blend
since this is the only time they're intended to be used.

Cleanup: replace 'timer_ended' with 'timer_end'

The previous naming made it seem as if the timer had already ended
where as this function ends the timers.

Cleanup: remove unused auto-save read function

Also make wm_autosave_location a static function.

Cleanup: refactor auto-save timer usage

- Split out auto-save file writing from timer code.
- Add wm_autosave_timer_begin so there are both begin & end functions.
- Replace WM_event_add_timer/WM_event_remove_timer with begin/end calls.

BLI: Add "is_zero" method to float2 and float3 types

This is not necessary, but a nice convenience to avoid using `is_zero_v3`.

Differential Revision: https://developer.blender.org/D10713

Nodes: multi-input support for Attribute Remove node

This patch adds multi-input support to the Attribute Remove node.

Reviewed By: Hans Goudey

Differential Revision: https://developer.blender.org/D10698

Fix T86427 Exact solver does not apply target material.

I had done some experiments to see what Fast boolean did for material
mapping and thought it just used the same slot in the target as the
slot in the source. The truth is more complicated: if the target material
exists in any slot of the destination, we need to remap to whatever
slot has the matching material. I fixed Exact Boolean to do this.
Since the materials may be in the object, this means that BKE_mesh_boolean
had to get another argument, the remapping arrays.

I will note that the current behavior of Fast, and now Exact, is not ideal.
Ideally, if the source material does not exist in the target, a new material
slot should be created in the target and the source material copied there
(and incrementing the material's reference count). Maybe a future project,
but for now, I want the behavior of Exact to match that of Fast.

EEVEE: Specular Occlusion: Avoid overdarkening on smooth surfaces

Accumulate error caused by the low amount of integration slices and
correct it for the low roughness surfaces.

This increases light leak but it is less distracting than dark fringe
everywhere.

CMake/guardedalloc: add header file to TEST_SRC

Tests: add script_validate_keymap

This checks the generated key-map data matches the result of
re-exporting and re-importing.

This shows up various inconsistencies, including:

- Unused keymaps.
- Unknown/unused data in the keymap.
- Event arguments that don't make sense.
- Event values that don't match the event type
(tweak direction on keyboard event for example).

Cleanup: remove unused "Gizmo" & "Header" keymaps

Cleanup: remove invalid repeat argument

Cleanup: remove redundant keymap arguments

Also add missing space-type argument, while not required
this is used by convention elsewhere in the key-map.

Cleanup: avoid risky list as default argument in keymap

Cleanup: unused argument warnings

Cleanup: use explicit check for WM_KEYMAP_UPDATE_RECONFIGURE

A block of code ran when `wm_keymap_update_flag` was non-zero,
replace this with explicit flag check since it wasn't
immediately obvious which flag needed to be set.

WM: support loading keymaps in background mode

While this still isn't done by default on startup,
activating a key-config will load it as expected.

Needed to perform key-map loading tests in background mode.