* added button to the bottom right to switch to higher contrast,
* suitable for projectors
* increased contrast between background and text colors
* thicker strokes around chart elements
* ensure yellow details panel backgrounds are darker
* added .btn-opacity to control opacity for buttons on hover/selected
state
* increased contrast in topology link labels
* removed dark background from status to not confuse with topo links
* tried activating it for child tables too, but that became quite CPU
intensive, and feed intervals became out of sync (host CPU < container
CPU, which is hard to believe)
* Refactored sparklines to be rendered by react
* Correct row key for sparklines
* Extracted data feed into animated-sparkline
* last value is rendered by sparkline now, because it relies on the
last value that it is fed, not the lastest availble value
The Chrome guys have implmeneted a new change (in Chrome 48) in the CSS
flexbox spec, min-width now defaults to auto instead of 0. "auto" has
some complicated behaviour described in the table here:
https://drafts.csswg.org/css-flexbox/#min-size-auto
This results in a very long min-width (I don't understand why). So we
have to let the browser know it can make it smaller if it wants to,
which in this case would be great.
Here's another person's experience with the issue.
- https://code.google.com/p/chromium/issues/detail?id=551336
- 'label_major' has been changed to 'label' in the details api response.
- getNodeColorDark was giving up too early if a rank wasn't provided.
The topo diagram which uses getNodeColor will select a color even w/
no rank. Hosts don't have ranks at the mo.
- Also fixes strange issue where hitting ESC while a terminal was
active, would update the route, reload the top card, and break the
terminal connection status.
- Treat control objects that come back from the server as little
black boxes.
- Pass our local client nodeId around more instead, use that for
comparisons etc, (vs. inspecting the control object and doing brittle
magic w/ the ids).
They stopped working because of the change to container node IDs (and rendered nodes IDs for containers). The UI was comparing the IDs, which was never safe. I have just removed that code. This does leave the possibility of us having the pipe control operation take a long time, the user navigate to a different node, and then the terminal pop up, but I think thats better that teaching the UI to understand the format of the node IDs.