search

cyoung / stratux

Aviation weather and traffic receiver based on RTL-SDR.
BSD 3-Clause "New" or "Revised" License
1.06k stars 365 forks source link

the Status websocket is reopenning after being explicitly closed #283

Open bradanlane opened 8 years ago

bradanlane commented 8 years ago

It turns out the resolution for #164 which was committed as 1c231dd has introduced a similar bug.

Now, when a user legitimately navigates away from the status page, the status websocket is reopened even though the status page is no longer current.

You can view the bug in the browser console. Start on the status page, and then navigate to the log page (or any other page). The console will continue to show Received status update

I'm not yet sure what the right mechanism is to differentiate between the socket being closed by the user vs otherwise. It may just be a flag.

ssokol commented 8 years ago

Not really a big deal but, would it be feasible to build this as a single page app with each of those tabs simply being a different view, rather than a different page? That way you could use a single websocket that connects when the page loads or resumes from background, and there would be no reconnect time when switching views.

On Tue, Feb 23, 2016 at 2:38 PM, bradanlane notifications@github.com wrote:

It turns out the resolution for #164 https://github.com/cyoung/stratux/issues/164 which was committed as 1c231dd https://github.com/cyoung/stratux/commit/1c231dd301c48c579250097f219dff382b98ad32#diff-709e310a9fdf1c8b5ad0d6d26d76e128 has introduced a similar bug.

Now, when a user legitimately navigates away from the status page, the status websocket is reopened even though the status page is no longer current.

You can view the bug in the browser console. Start on the status page, and then navigate to the log page (or any other page). The console will continue to show Received status update

I'm not yet sure what the right mechanism is to differentiate between the socket being closed by the user vs otherwise. It may just be a flag.

— Reply to this email directly or view it on GitHub https://github.com/cyoung/stratux/issues/283.

Steven Sokol 408 Camelot Drive Liberty, MO 64068

mobile: +1 816-806-8844 fax: +1 816-817-0441

bradanlane commented 8 years ago

I'm by no means an AngularJS guru so I would not know what pitfalls will come from trying to merge everything at this stage.

I removed the timeout line in status.js and made all files served from managementinterface.go with no-cache, no-store. Thus far, my testing is showing not bad side effects.

cyoung commented 8 years ago

What do you suggest to fix this?

We need a similar fix for #255.

bradanlane commented 8 years ago

I'll need to attempt reproducing the Firefox+Ubuntu case to see if it's the same. In this latter case it's not a production configuration so it may be less urgent. Hopefully the fix for #283 and #255 will be the same/similar.

My above mentioned fix works for #283 but the no-cache no-store is a bit of a hammer-kludge and should be done "proper" for performance reasons.

Ergonomicmike commented 8 years ago

I don't know where a performance issue would manifest itself, and I know it's more elegant to use elegant solutions instead of kludges. But to put performance in perspective, I only see 30% ave CPU load in flight. And that's with the powersave gov loafing at 600 MHz.So if a hammer bumped that up to 50%, there's still plenty of headroom.

bradanlane commented 8 years ago

OK. I'll create a branch with my "fix" and submit the pull request.

Ergonomicmike commented 8 years ago

Worth a try. Still in beta, after all.

westphae commented 7 years ago

Old issue, but I noticed that the gps socket persists after the page is changed too. Both status and gps sockets only persist if the user stays on that page for a few minutes before switching. Interestingly, I have been unable to get the traffic socket to persist. Its onclose event is handled a bit differently, but modifying the status and gps sockets' onclose methods to be like traffic's didn't seem to resolve the issue. But maybe the traffic socket's better behavior is a clue to resolving.