This is a potential breaking change for behaviors which
rely on being able to change content in-place even if
the response's content is not applied. This is the case
for grafana's iframe behavior, though that's more a hack
than a necessity so it doesn't break. I've found no other
implementations so I guess it's fine including it here.
A redirect caused by a form submit is **never** an automatic submit.
Unless I'm missing something here. There is one way this may be true,
as ipl forms without a submit button are indeed successful due to a
autosubmit element, but they don't redirect by default. So if the
controller redirects in such a case, the form should have a submit
button, I guess..
Anyway, this is necessary due to the previous commit as this may
otherwise cause form submits, that re-render layout during a redirect,
are considered being auto submitted. (Such as the login form)
- Previously, in case of `abort` the url was already pushed to history at the beginning of `onFailure`, setting `addToHistory` to true only prevents it from being added twice (again in `onComplete`).
- Now we only push the url to history in `onComplete`, so setting `addToHistory` to false is only necessary if req is scripted.
- We intentionally push the canceled url into history to continue navigating with the browser`s back/forward key.
Not the container's base target. The redirect target may still be
overridden by `data-base-target` on the `form` or by the response
using `X-Icinga-Container`.
* Let response header `X-Icinga-Container` either influence a request's `$target` or `$redirectTarget`, but not both
* Close a request's `$target` instead of the `$redirectTarget` upon the redirect url `__CLOSE__`
* Issuing a `__CLOSE__` in a detail url still closes `#col2` and refreshes `#col1`
* Issuing a `__CLOSE__` for a nested container still empties it
* Issuing a `__CLOSE__` in a modal, now refreshes the modal openers container
* If `X-Icinga-Extra-Updates` is set, automatic refreshing won't happen
Search requests are handled elsewhere in a more dynamic
way since .. some time. This part also used to be active
only in case Icinga Web is hosted at the root of the web
server, which many users don't do as otherwise we would
have gotten more reports in the past regarding the reload
animation. Or rather, about the login not working due to
the same reason.
It's gone for good now.
fixes#4834
..if a renderHook discards an autorefresh. There is
unfortunately no way to distinguish between a renderHook
that really discards changes or one that applies them
on the DOM itself. If it's the first, this change
*shouldn't* hurt. If it's the latter, users should
benefit.
This has the same effect as a normal redirect. The benefit of it however
is that the server doesn't need to know what's being shown in the left
column. It just instructs the client to close the right and refresh the
left column. But still produces a new history state, it's a forward
navigation nonetheless.
Isn't the prettiest fix. I would have liked to completely
remove the target preparation from `getLinkTargetFor`.
But this is the easiest fix since it's only for modals
that preparation is not desired. It's also the most
compatible change.
