STR:
(1) Apply bug 1580012's patch and set layout.css.xul-box-display-values.content.enabled to true [EDIT: fixed bug number]

(2) View attached testcase.

ACTUAL RESULTS:
The second item (the display:-moz-box one) is only as wide as its contents.

EXPECTED RESULTS:
That item should stretch to the full width of the grid (the default justify-self:stretch behavior).

Also, the last line of the testcase should report "[-moz-box]" -- if it says [block], then that's a hint that you didn't finish STR step (1).

It turns out this happens because we pass in ComputeSizeFlags::eShrinkWrap to ReflowInput's ComputeSize() call for this frame.

We do that because mFrameType (which has a version of the "display-outside" for the frame) is not block-flavored for this frame, and that means we end up initializing the computeSizeFlags to eShrinkWrap instead of eDefault here:
https://searchfox.org/mozilla-central/rev/588814f2edddf0e132d77d326ddae50911e8bad1/layout/generic/ReflowInput.cpp#2382-2385

      bool isBlock = NS_CSS_FRAME_TYPE_BLOCK == NS_FRAME_GET_TYPE(mFrameType);
      typedef nsIFrame::ComputeSizeFlags ComputeSizeFlags;
      ComputeSizeFlags computeSizeFlags =
          isBlock ? ComputeSizeFlags::eDefault : ComputeSizeFlags::eShrinkWrap;

If I add an exception to this for grid-item box frames, then I get EXPECTED RESULTS.

Note that if we changed XUL box frames to have DisplayOutside() report Block (emilio's approach in the try patches on bug 1580012), then this would Just Work per this chunk, I think:

    switch (disp->DisplayOutside()) {
      case StyleDisplayOutside::Block:
      case StyleDisplayOutside::TableCaption:
        frameType = NS_CSS_FRAME_TYPE_BLOCK;
        break;

https://searchfox.org/mozilla-central/rev/588814f2edddf0e132d77d326ddae50911e8bad1/layout/generic/ReflowInput.cpp#899-902

Here's a simple patch that should fix this (when combined with my initial patch in bug 1580012). Just posting to unblock ntim's experimentation for the moment.

As noted in prior comment: if we take a different approach on bug 1580012, then we may not need any additional patch.

I'm assuming this is going to be blocking the frontend work to remove XUL grids tracked in Bug 1520625.

This is with:

https://hg.mozilla.org/try/rev/3686e153c092316ab975a7b383752fb86c51b814
https://hg.mozilla.org/try/rev/8048fc847b3927facdc5d4df673b63ea859ba000

applied

Also: rename "isBlock" to "isBlockLevel" to reflect reality. Its naming is
based on the constant NS_CSS_FRAME_TYPE_BLOCK, which in fact is not specific to
display:block but in fact is for frames that are "block-level in normal flow"
(which I think in practice means block-level and not out-of-flow).

https://searchfox.org/mozilla-central/rev/7531325c8660cfa61bf71725f83501028178cbb9/layout/generic/ReflowInput.h#51

https://hg.mozilla.org/mozilla-central/rev/088c878f9f7c