[css-position-3] A bug about "weaker-inset" of inset-modified containing block

[xiaochengh]

The spec defines "weaker inset" as:

In the case that only one inset is auto, that is the weaker inset; otherwise the weaker inset is the inset of the end edge (where end is interpreted relative to the writing mode of the containing block).

And if the previous steps of IMCB gives a negative size in any axis, the weaker inset is reduced to bring the size back up to zero.

This seems incorrect when both insets are auto, in which case the IMCB is produced by extending from the static position. So the weaker-inset-reduction process should be a reversal of that extension. So instead of looking at the writing modes and directions, it should be:

• For self-start alignment or its equivalent, the weaker inset is the end inset (in the element's own writing direction, ditto below)
• For self-end alignment or its equivalent, the weaker inset is the start inset
• For center alignment, both insets are equally weak, and should both be reduced by the same amount until they meet at the static position

@tabatkins