Fix wrong GL DepthTest param

[Helwor]

That may disturb other widgets (especially Eco Drawer).

[GoogleFrog]

Have you checked whether the widget that runs after this one (by default) happens to rely on this state being left there?

[GoogleFrog]

I ask because either you've tested it, or I am going to have to.

[sprunk]

widget:DrawGroundDeferred() doesn't seem to be a real thing. Maybe it's BAR specific?

The full list of widgets using widget:DrawWorldPreUnit() after enabling this one:

Mex Placement Handler
Custom Markers
Lups
AllyCursors
Map Edge Extension
Commands FX
Map Edge Extension 2        <----
Showeco and Grid Drawer
Missile Impact Points
Defense Range Zero-K
Decloak Range
Antinuke Coverage
Selected Units GL4 2
API Unit Tracker GL4

It looks like lots of things below and above happens to rely on whatever state they inherit from the previous widget. I put a little test on the glDepthTest-debug branch and at least defense ranges, antinuke ranges, mex spots, and OD pylon circles get broken in various ways if they start with the wrong glDepthTest value.

[Helwor]

Yes I've tested it.
As a widget, when you says gl.DepthTest(true) you expect it will suppress anything below ground.
For that it need the param to be set to GL.LEQUAL (or GL.LESS but it's not default). So having it as GL.ALWAYS is obviously problematic. It basically means that a widget doing gl.DepthTest(true) will draw as same as gl.DepthTest(false) aka draw everything. Widgets never expect that param to be changed. A widget that changes it should mandatorily reset it at the end.
Letting DepthTest ON or OFF, is a bit less problematic bc usually a widget will explicitly say if he want to draw with depth constraints or not. From what I understood: DrawWorldPreUnit have DepthTest false by default and DrawWorld have it true, not sure 100%.
Letting Culling ON is problematic bc a widget usually don't expect the culling to be activated and don't bother set it to false before drawing.
Letting Culling Param is less problematic bc a widget using culling will most likely explicitely state which face he want to cull.
Eco Drawer use a shadow volume stencil trick to draw circles perfectly on the ground (glVolumes.lua). For that it need culling to be false and expect depth test to be GL.LEQUAL when activated.
Idk about other widgets, but if they are disturbed by that change (depth param GL.LEQUAL, culling false), they should adapt imo. For the reasons cited above...

[Helwor]

I havent tested it yet, but widget:DrawGroundDeferred() is probably relying on Spring Settings
AllowDrawMapPostDeferredEvents
AllowDrawMapDeferredEvents
AlwaysSendDrawGroundEvents

[sprunk]

Idk about other widgets, but if they are disturbed by that change (depth param GL.LEQUAL, culling false), they should adapt imo.

Yes, what GF was asking for was that part of this PR should be checking if this is the case for the default set of enabled widgets and adapting anything that gets disturbed, otherwise we will still have bad rendering and are just shuffling it around between different widgets.

[GoogleFrog]

Exactly. I'm not saying "we must keep broken things". I'm saying that if fixing something that is technically broken causes other things to be predictably broken, then the PR with the fix should also fix the new practical brokenness.

[Helwor]

I haven't tested thoroughly, I just never have this widget ON, and this widget is OFF by default, therefore if fixing it would break next widgets, next widgets would also be broken just by the fact this widget is OFF... simple logic. Context is that I had a user comlpaining about eco grid not showing properly and I found why.