cannot destroy a map with panels

[erilem]

This is a regression introduced by [10732], where destroying a panel that includes (already) destroyed controls produces an error. This is critical because the bug triggers when destroying a map that includes a panel.

[erilem]

test demonstrating the bug

[fredj]

looks good, please commit

[jorix]

Had yet to send the same bug, I think it is better to change "destroy" from "control.js" putting:
this.active = null.
This rearrange future incidents similar.

[erilem]

Replying to jorix:

Had yet to send the same bug, I think it is better to change "destroy" from "control.js" putting:
this.active = null.
This rearrange future incidents similar.

jorix, please provide a patch, I'm failing to understand what you're suggesting. Thanks.

[jorix]

SVN 10810

[jorix]

Replying to erilem:

Replying to jorix: ... jorix, please provide a patch, I'm failing to understand what you're suggesting. Thanks.

I wrote fast and my English is not good.

I added patch based on r10810.
Becomes "this.active = null;" in the destruction of control.

I have seen that there are some controls that are deactivated by the destruction and there is comment

this.deactivate(); // TODO: this should be handled by the super

(see Split.js and Snapping.js)
I understand that in the future in the destruction of a control is also deactivated. For this reason and because it only runs once I think better this option.

[jorix]

there is another reason, the test:

control.activate();
control.destroy();
t.eq(control.deactivate(), false, "control destroyed can not be deactivated.");

fails.

---

And I think it would be nice if

t.eq(control.activate(), false, "control destroyed can not be activated.");

did not cause an error, as now, but this would be another patch.

[ahocevar]

r10810

[ahocevar]

r10810 does not resolve the issue. I'm looking at the patch that jorix has provided now.

[ahocevar]

jorix's patch also does not solve the issue. I am investigating this now, but if it takes more than 15 minutes to find a solution, then we should really consider to revert r10732.

[ahocevar]

The above patch reverts r10810 and partially reverts r10732. It fixes the deactivate problem of #2835 by calling redraw. The remaining destroy issue in #2835 is a duplicate of #2210 and should be handled there.

All tests still pass. Thanks for any review.

[erilem]

This totally makes sense. Please commit.

[ahocevar]

r10834