There were two issues which make the test fail.

1. We do descend down descendant elements inside a scroll container even if we meet other scroll containers there. The spec clearly says at the last sentence of Scroll Snap Model section that is that "Snap positions only affect the nearest ancestor scroll container on the element’s containing block chain"

2. In this commit I did add code that makes a choice whether the given range is passed to ScrollToWithOrigin (or ScrollTo) or not to not constrain the final destination position modified by snapping. But it was not sufficient since ScrollToShowRect in nsPresShell also calls ScrollTo with an nsRange. What we should do instead is in ScrollToWithOrigin if we found a valid snap position in a call of GetSnapPointForDestination we ignore the given nsRange there.

https://treeherder.mozilla.org/#/jobs?repo=try&revision=021985ca37393eda280b3804380fdf14a381eee2

From the last sentence of "3. Scroll Snap Model" [1] in the spec;

Snap positions only affect the nearest ancestor scroll container on the
element's containing block chain.

[1] https://drafts.csswg.org/css-scroll-snap-1/#overview

ScrollToShowRect calls nsIScrollableFrame::ScrollTo with an nsRange which
will be used to constrain the final scroll position so that if snapping needs
to happen we need to ignore the given range not to constrain the final
destination position in the range.

Depends on D31947

Backed out 2 changesets (bug 1553022) for build bustage on builds/worker/workspace/build/src/layout/generic/nsGfxScrollFrame.cpp

Log:
https://treeherder.mozilla.org/logviewer.html#/jobs?job_id=247928664&repo=autoland&lineNumber=32549

Push with failures:
https://treeherder.mozilla.org/#/jobs?repo=autoland&revision=d136380e4945c8081a2952760143a0bb1e36df7c

Backout:
https://hg.mozilla.org/integration/autoland/rev/97fd39fe7f582ff714a56a65194bf9ca01711686

gcc isn't smart enough, I'd say. :/

https://hg.mozilla.org/mozilla-central/rev/f68ba99112d6
https://hg.mozilla.org/mozilla-central/rev/e882a5c576ab