Bugzilla

In bug 1631941 I fixed some inconsistency in our code, and switched to treating the borders and padding that are skipped due to fragmentation as always being in the frame's writing mode.  This was what most, but not all, of our code did at the time.  (There were only two callsites that [needed to change](https://hg.mozilla.org/mozilla-central/rev/f4c1d7dcca50), and for one of those the change only affected a small portion of *its* callers.)

However, looking at the spec, this behavior is actually incorrect.  The spec actually [defines specific rules for this](https://drafts.csswg.org/css-break-3/#joining-boxes):

* for fragments across lines, the spec's normative text is inconsistent with the examples, so I'm not sure
* fragments across columns use the multi-column container's direction (and analogous for regions), so our current code is wrong
* fragments across pages use the root element's block flow direction, so our current code is also wrong

In bug 1631941 I fixed some inconsistency in our code, and switched to treating the borders and padding that are skipped due to fragmentation as always being in the frame's writing mode.  This was what most, but not all, of our code did at the time.  (There were only two callsites that [needed to change](https://hg.mozilla.org/mozilla-central/rev/91ac0b7e571bbee079c5040f5c8fb24bf44eca7e), and for one of those the change only affected a small portion of *its* callers.)

However, looking at the spec, this behavior is actually incorrect.  The spec actually [defines specific rules for this](https://drafts.csswg.org/css-break-3/#joining-boxes):

* for fragments across lines, the spec's normative text is inconsistent with the examples, so I'm not sure
* fragments across columns use the multi-column container's direction (and analogous for regions), so our current code is wrong
* fragments across pages use the root element's block flow direction, so our current code is also wrong

In bug 1631941 I fixed some inconsistency in our code, and switched to treating the borders and padding that are skipped due to fragmentation as always being in the frame's writing mode.  This was what most, but not all, of our code did at the time.  (There were only two callsites that [needed to change](https://hg.mozilla.org/mozilla-central/rev/6120b8d2df59), and for one of those the change only affected a small portion of *its* callers.)

However, looking at the spec, this behavior is actually incorrect.  The spec actually [defines specific rules for this](https://drafts.csswg.org/css-break-3/#joining-boxes):

* for fragments across lines, the spec's normative text is inconsistent with the examples, so I'm not sure
* fragments across columns use the multi-column container's direction (and analogous for regions), so our current code is wrong
* fragments across pages use the root element's block flow direction, so our current code is also wrong