Also, the RESTRICT_INSTANCED_OFF icon looks like thicker version of the linked icon. It and its RESTRICT_INSTANCED_ON counterpart are not used.

Ah, rB777a546c0c7 seems like those were designed for before the "restrict viewport" was a thing, they never even reached 2.80.

In D13081#373967, @Campbell Barton (campbellbarton) wrote:

Noticed some additional issues, suggest fewer changes P2782.
@Alessio Monti di Sopra (a.monti) do you see any issues with this change? If not, I'll update the patch and commit this.

• removed the ui reordering part, and placed the property inside the existing IMAGE_PT_overlay_image panel

• added the missing bit to make the highlighting work when hovering overlapped buttons in uiLists

Actually, it's probably better to just update the patch with that mentioned change included, so that it's clearer what I mean.

The above (roughly) matches what is happening for ui_handle_tree_hover, but a bit simpler. But still didn't work. LOL.

No idea.. What I did instead is to expand that function to include both the list and tree cases while keeping the same logic and, it doesn't work.
The only thing I noticed in debug is that the function gets executed a lot of times in a row with different "events" and slightly different cursor positions. Not sure if that is expected, as I'm definitely not familiar with that.

In D13858#370369, @Julian Eisel (Severin) wrote:

• This needs versioning code so the highlight color works for preferences saved in earlier versions, see do_versions_theme().

• update to the current state of master

Bring back the new UI_BUT_ALIGN_IS_FIRST flag.

Simplifed and moved the logic away from ui_text_icon_width_ex()

Update to current state of master

Hi, have you considered doing something similar to what we have in the video sequencer for the toggle placement?

• make variables constants

@Yevgeny Makarov (jenkm) It seems logical that we should change buttons coordinates somewhere in one place. The question is where.

One place could be just after ui_block_align_calc() in interface.c. If we could do it there, a simple function that moves all the buttons to the left and then the block half way to the right might work, but I don't know if it's possible to check for the layout flags there.

• cleanup
• comments

In D13156#355256, @Hans Goudey (HooglyBoogly) wrote:

This looks reasonable, I just wonder if adding a new widget type is the best way to handle that. I don't really know what the alternative would like like, but maybe a check in the existing UI_WTYPE_ICON?

Better comments

@Erindale (Erindale) if you pin the cursor panel then this gets quite fast, you still need to go to the 3d view to place the cursor, so at worst you loose a couple of seconds to press N and scroll to the top of the panels, if you're not already on the top; (providing that copy/pasting the whole vector becomes the default with ctrl+c on hover, and that this gets properly publicised in the release notes).

You can with Alt-Ctrl-C / Alt-Ctrl-V.

Isn't the poll a bit too loose? This gets in a lot of places.
Why would someone want to paste the cursor loc/rot in a scale property, the SSS radius, a color socket, and so on?

• Rebase on current master
• Removed .dat files for force_boid icon

Try to use arc to get it to work

In b.chat there was also a discussion about having an svg file icon, now that they're getting more integrated thanks to GP (right now they are using the generic FILE_3D, same as obj and other IO formats).

@Richard Antalik (ISS) could you confirm that this solves the problem without causing other ones? It seems so, but I'm not sure it should.

diff --git a/release/scripts/presets/keyconfig/keymap_data/blender_default.py b/release/scripts/presets/keyconfig/keymap_data/blender_default.py
index dbc93cb6caa..aa20b5b0396 100644
--- a/release/scripts/presets/keyconfig/keymap_data/blender_default.py
+++ b/release/scripts/presets/keyconfig/keymap_data/blender_default.py
@@ -7557,7 +7557,7 @@ def km_sequencer_editor_tool_generic_select(params, *, fallback):
                 params, "sequencer.select", "sequencer.cursor_set", extend="toggle")),

Thanks for the quick triaging @Pratik Borhade (PratikPB2123) !

• simpler changes, adding bContext wasn't needed
• fix action zones being too small with region overlap

• check for G_TRANSFORM_CURSOR in the drawing poll function
• renamed the new cursor flag to SI_OVERLAY_HIDE_CURSOR and move it in the eSpaceImageOverlay_Flag enum
• moved the rna definition in rna_def_space_image_overlay() and renamed the flag "show_cursor"

To complete this, a secondary patch I wanted to work on would be about blocking the cursor position while it is not drawn. I initially though to include it here for the image editor, but I think that should affect all the editors that have a cursor.

@Christoph Lendenfeld (ChrisLend) It fails to build, the structs used in the functions are defined below them.

In D12977#341726, @Harley Acheson (harley) wrote:

@Alessio Monti di Sopra (a.monti) - ...GCC gives a lot of conversion errors...

Thanks! I think those should all be gone now. I wish our compilers gave the same warnings/errors.

Hi @Harley Acheson (harley) I wanted to try this but GCC gives a lot of conversion errors related to those FreeType custom types.

Well, ok, but to me it still seems quite a random thing to do.
I mean, I see no gain in hiding the icon, and the blank space looks a bit strange.

Hi! I honestly don't get why the icon couldn't be always visible.

In T76023#916799, @William Reynish (billreynish) wrote:

Part of the idea here is that you usually only use a few categories at a time. You could move those to the top. And collapsed you could of course open and switch to any category.

• addressed inline comments

In future please don't mix code cleanup with functional changes, mixing in these changes slows down code review.

Sorry about that, I hoped it was clear enough from the patch description.

Hi! I was trying the commit and I noticed a couple of problems. I thought it was better to just list them here instead of filling bug reports.

@Sybren A. Stüvel (sybren) @Luciano Muñoz Sessarego (looch) I was looking at the group colours thing, do you wanted the change to only affect the Graph editor or also the Dopesheet?

• moved adding vertices in separate function
• used floor(pixelx) for step values higher than 1
• changed max_verts calculation
• inverted logic for precondition

• moved condition to change_frame_seq_preview_begin() using ED_time_scrub_event_in_region()

In D7234#173045, @Julian Eisel (Severin) wrote:

This effectively disables the preview feature entirely for the LCS keymap. If you guys say that's fine, sure go ahead, but it should be explicitly mentioned in the commit log.

On the code side - shouldn't the caller handle this logic? This function should simply and only set the preview to whatever strip is closest, after all it's called _set(). The caller should only execute that when it makes sense. You can also use ED_time_scrub_event_in_region() there then.

• addressed inline comments
  • added preconditions
  • changed variable names
• get seq->start/end from the main function

I have done comparison of batch vs immediate drawing, and there is little to no improvement in performance. I did not expect that...

• further reduce resolution of the curve at lower zoom level
• add the last couple of vertices out of the main loop, in order to make sure that the shape is always properly closed

• moved evaluation steps to integers
• add an extra evaluation frame left/right of the editor's borders
• avoid adding verts for points next to each others that have the same value

• split code for invalid and locked states in separate functions
• same for code calculating text x-axis offsets

• small fix to the text margin, now it considers UI scale
• use that margin also for missing media indication

• better vertical position for the text, allowing to show more of the strip content (color bands, waveforms) when the strip is wider.

• forgot to update color of trim numbers after UI_view2d_text_cache_add() changes
• forgot to remove part of f-curves drawing related changes

• removed previously added header that wasn't needed
• removed f-curves drawing function, will try to implement that properly in a separate patch