(fix) - scu without children

[JoviDeCroock]

No description provided.

[coveralls]

Coverage increased (+0.1%) to 99.881% when pulling 9a8859c on fix/noSiblings into 7345fe8 on master.

[JoviDeCroock]

Hey @mzealey if you feel up to it, feel free to test with this branch. I think all cases SHOULD be resolved. Thank you so much for the error reports and following this up

[mzealey]

Yes this fixes my live use-case thank you!

[andrewiggins]

Nice work! Thanks for the good test cases 👍

[sventschui]

When a component bails out using sCU===false it's children are not diffed. This lead to outdated _parent pointers on the children what ultimately lead to updateParentDomPointers update the _dom pointer of a wrong/old parent vnode. Thanks to @JoviDeCroock for the support in finding this.

My fix in #229c135 fixes this. It adds some bytes to the bundle (Update: Travis says it adds 32 bytes). Since I couldn't get my local build to run (terser issue) I could not yet do advanced bytes golfing, so feel free to try your luck reducing the footprint of this fix 🍀

[marvinhagemeister]

So good! Love that you two worked together to get to the bottom of it 👍 💯

[sventschui]

for (tmp = 0; tmp < newVNode._children.length; tmp++) {
	if (newVNode._children[tmp]) newVNode._children[tmp]._parent = newVNode;
}

vs.

newVNode._children.some(child => {
	if (child) child._parent = newVNode;
});

where the later saves 11B but is slower during runtime. What to prefer?

[developit]

@sventschui my suggestion would be that we use the for loop for now, but we should do a PR to convert not only this but other loops to other syntax at the same time - that will let us measure the effect without it being confounded by mixed syntax.

Also I wondered if this was smaller?

const c = newVNode._children;
for (tmp = c.length; tmp--; ) {
    if (c[tmp]) c[tmp]._parent = newVNode;
}