This spun from bug 137367 comment 12 and from bug 137367 comment 15 .

Tests

http://www.gtalbot.org/BrowserBugsSection/CSS3Multi-Columns/columnfill-auto-widows-001.html

http://www.gtalbot.org/BrowserBugsSection/CSS3Multi-Columns/columnfill-balance-widows-001.html

Reference for columnfill-auto-widows-001.html

http://www.gtalbot.org/BrowserBugsSection/CSS3Multi-Columns/reference/columnfill-auto-widows-001-ref.html

Tests fail in Firefox 60.7.1 ESR and in Firefox 69.0a1 buildID= 20190629100032 .

Tests pass in Chromium 73.0.3683.75, in Chromium 77.0.3839.0 and in Epiphany Tehcnological Preview build 3.31.90-8061c1929 (WebKitGTK 2.25.2). Tests presumably also pass in Opera 12.16 (Presto engine).

• I intend to submit those tests to the web platform repository
• The tests will then eventually be slightly modified or adjusted so that they do not require a viewport wider than 600px
• The test columnfill-balance-widows-001.html may become a manual test as I am not sure I can create a reliable and trustworthy reference file for it
• I searched for duplicates and did not find any.

The div#reference in columnfill-auto-widows-001.html test could also be some kind of grid or 3 inline-blocks side-by-side instead of a multicolumn element. I will try to achieve this, when I have more time.

I just checked with Browsershots.org and Opera 12.16, indeed, pass both tests.

I don't see why this is useful to have as a separate bug; I don't see any reason it would be easier to do for multicolumn than it would be for pages. It might actually be harder for multicolumn (due to having to re-layout), but a separate bug shouldn't be filed for that unless the person implementing the feature for pages decides not to implement it for multicolumn.