Removing the "." that was in {,.[0-9]} allows it to match /dev/cuau10 and onward.
I added lots of comments on the glob expression, because the format of the glob expression is not the same as an ordinary regex.
If there is an input error then the edit page is redrawn showing the
input errors. The HTTP_REFERER becomes the current page, rather than the
true original referer. Then if you click Cancel the current page is just
redrawn.
This change makes the code remember the original referer, so if the user
enters some invalid data, presses Save, reads the input error messages
then presses Cancel they are taken back out to the original page - the
same as if Cancel was used before Save.
When adding a Virtual Server, if you press Save with blank fields, the validation does not show. That was because the switch statement selecting the fields to validate had wrong case values. Actually the "mode" here is locked to "redirect_mode" (later code to allow the mode to be selected by the user is commented out - not implemented.
I fixed the reqdfieldsn array also for relay_mode case, even though it is never used. It looked dodgy the way it was.
If you go to Interfaces, assign, PPPs, press "+" to add an entry, then press Save without entering anything then you get a warning about the foreach() here.
The is_array() check fixes it.
When enabling or disabling a route by using the enable/disable button on the Routes page, the destination network was not being put in the config change description, although the code intended it to be there.
This fixes it.
The Traffic Graphs widget puts a graph object for every interface into
the HTML of the widget. Underneath the graph object for every interface
makes the call to graph.php to get data for the graph, and the refresh
interval also keeps happening, which keeps gathering the data every
refresh interval for every interface.
This wastes a lot of CPU back on the pfSense box gathering data
repeatedly if the system has a lot of interfaces and only has the
Traffic Graph enabled for 1 or a few of them. e.g. on my poor little
Alix at home I had setup 6 VLANs to test some stuff, so I had WAN, LAN,
OPT1 and 6 tagged VLANS - 9 interfaces. When I enable the Traffic Graphs
widget on the dashboard, the Alix CPU is driven 100% trying to keep up
with gathering data for 9 interfaces every 10 seconds, even though I
only have 1 graph actually opened!
I couldn't see a way to enable/disable the graph.php object from
executing. So this change puts the object in and out of the HTML:
a) At first display the graph.php object HTML is only put in the
relevant "div" if the graph is actually set to be shown in the config.
b) When a graph is opened by the user, the Java Script puts the
necessary graph.php object HTML into the div. The graph data then starts
being requested...
c) When a graph is minimised by the user, the Java Script removes the
graph.php object HTML from the div. The actions of graph.php stop
happening.
I can see the difference by just watching "top" from the command line on
a low-power system and open/minimizing the graphs on the dashboard.
If there is a better way to enable/disable an object like this I am
happy to learn about it.
when the page is first displayed.
This has annoyed me a few times and it annoyed me again just now. I had some settings in one of the advanced boxes that were no longer wanted, but when the services_dhcp page is shown the contents of advanced settings are not shown to the user - the user has to click on each advanced button to see if there are any settings already there.
This change displays the settings automatically, if there is something already specified. It works the same way as the Firewall Rules advanced boxes.
The log and picture widgets were both using "iforma" and "submita".
Actually it did not break anything because it was only the form name
repeated, not id. And nothing was using these names.
Traffic Graphs widget was using just "iform". That is a bit dangerous
for the future. I got tricked when cut-pasting some code to make some
settings options for the Gateways widget. I kept "iform" and then
wondered for a while why my Traffic Graphs widget show-hide settings
would not save. There was traffic graph JS that referred to just "iform"
and that started modifying the "iform" in my new Gateways widget code.
Rather than having names "iforma", "iformb"... "submita", "submitb"...
it seems much less risk of accidental duplication if these are named
like:
name_of_widget_iform
name_of_widget_submit
I don't think there is any user-visible bug in 2.2.* - so this
standardization could just go into 2.3
among all widget forms.
Traffic Graphs widget already uses the vanilla name "iform". Reusing that name causes Traffic Graphs widget graph display state (show/hide) not to be saved if GW widget was also displayed on the dashboard.
This fixes it.
since we can sometimes provide a useful description from that config
data also.
Fill the $iplookup array with host or FQDN data if description is blank
or host or FQDN was requested. Then we can use $iplookup in all cases
where we have local data. It simplifies some logic a bit.
This change adds a setting for the Gateways dashboard widget so the user
can choose to display the Gateway IP, Monitor IP or both.
If "both" is chosen and the Gateway IP is the Monitor IP, then only the
Gateway IP is shown - i.e. the same IP address is not repeated on the
widget display.
If "both" is chosen and the Monitor IP is different to the Gateway IP
then the Monitor IP is shown in () brackets after the Gateway IP.
If "Monitor IP" is chosen and there is no special Monitor IP defined,
then the Gateway IP is displayed (which is also the Monitor IP).
If "Gateway IP" is chosen then the widget behaves as it does now.
"Gateway IP" is the default.
I find this handy because the Gateways widget reports RTT (latency) and
loss figures that are actually for pings to the Monitor IP. So it seems
useful to be able to display the Monitor IP in the widget.
At the confirmation dialog after pressing the "Reinstall XML" button, the text does not distinguish between having pressed "Reinstall the whole package" and "Reinstall the GUI/XML". It would be nice if the text of this confirmation allowed the user to be confident about which button they had just pushed, before confirming the action.
Note: This stuff has no gettext() wrappers - but that can be fixed later, not get mixed up in this.
I thought that "reinstallxml" should do less than "reinstallpkg" but actually it was getting stuff here, then falling through "reinstalpkg" which did delete_package_xml and then install_pkg, which got the files a 2nd time and...
Maybe that was supposed to happen?
Anyway, I thought I would point this out and someone can either commit this pull request if the "break" should be there, or explain to me why "reinstallxml" is supposed to fall through executing all this code.
