821210 - reftest takes low resolution screen shots when run on a high DPI display

Status: NEW

(Testing :: Reftest, defect)

Description

[Cameron McCormack (:heycam)]

Running reftests locally on my MBP with its high DPI display, the test/reference images seem to be low resolution versions of the actual window contents.  It would be good if we could output 2x screen shots from the reftest harness and for the reftest analyzer to show them.

Comment 1

[Mark Smith [:mcs]]

This possibly depends on bug 780362.

Comment 2

[Cameron McCormack (:heycam)]

Attachment: patch

This makes the reftest harness capture native resolution images and updates the reftest analyzer to handle them.  The analyzer is updated to have this behaviour:

* When viewing a standard DPI reftest log on a standard DPI machine, no change.
* When viewing a HiDPI reftest log on a standard DPI machine, the reftest images
  are scaled down to 800x1000 and you can only inspect one in every four pixels
  by moving the mouse over the image.
* When viewing a standard DPI reftest log on a HiDPI machine, no change;
  the reftest images are shown at 800x1000 CSS pixels, and you can inspect
  every pixel of the original image.
* When viewing a HiDPI reftest log on a HiDPI machine, the reftest images are
  shown scaled to 800x1000 CSS pixels, which for a 2x log and machine, will
  result in the images being shown at their native resolution.  You can only
  inspect one in every four pixels with the mouse.

It probably makes sense to allow displaying the images at 1dpx = 1 CSS px so that you can inspect each pixel value, but I haven't done that for the moment.  This also probably only works well for integer values of layout.css.devPixelsPerPx.

Comment 3

[Cameron McCormack (:heycam)]

I didn't seem to need the new high DPI backing store canvas functions btw; I just sized them to the device pixel size I wanted and scaled the context by devPixelsPerPx before calling drawWindow().

Comment 4

[Timothy Nikkel (:tnikkel)]

Comment on attachment 715245 [details] [diff] [review]
patch

This appears to be a reftest log and not a patch.

Comment 5

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

Comment on attachment 715245 [details] [diff] [review]
patch

Indeed.  Sorry for not looking at this sooner.

Comment 6

[Cameron McCormack (:heycam)]

Attachment: patch

The actual patch this time.

Comment 7

[Cameron McCormack (:heycam)]

Comment on attachment 728779 [details] [diff] [review]
patch

A new patch will be needed.