Fix A8-on-R8 swizzle test on Chrome

[lexaknyazev]

Fixes this test on Chrome due to #2818.

[kdashg]

gl_PointSize is only valid when positive (and within ALIASED_POINT_SIZE_RANGE), so setting gl_PointSize shouldn't matter here, as it would be unsafe for WebGL to treat these as undefined.

WebGL doesn't specify what happens if gl_PointSize is unset, but gl_PointSize should always be being clamped to within the point size range, which does not allow zero.

Also, transform feedback with needs to run on drawArrays(POINTS) regardless of gl_Position or gl_PointSize, so skipping the draw (which is what it sounds like is happening?) here would be invalid.

[lexaknyazev]

I think that this particular test (active vs bound texture) shouldn't depend on an unrelated implementation issue that could be resolved separately.

[lexaknyazev]

gl_PointSize should always be being clamped to within the point size range, which does not allow zero.

GLES doesn't suggest that (from 3.0 section 3.4):

If the value written to gl_PointSize is less than or equal to zero, or if no value is written, the point size is undefined.

[kdashg]

I filed #2822 to add an explicit test for this.
Curiously, with transform-feedback enabled, chrome passes this test.
This sounds a lot like Chrome is being over-eager to skip draw calls.

[kdashg]

Let's talk about this in that PR, but I generally don't like working around implementation bugs in unrelated tests.

[kdashg]

To keep coverage, I would like to resolve #2822 before merging this.

[kenrussell]

This test was not intended to test the undefined behavior of failing to write gl_PointSize in a vertex shader used to draw points. Let's continue to discuss that in #2818 and #2822 . Merging this now.