Last time I was doing optimizations (maybe a year ago), I was able to see all possible function calls in the js flamechart.
Now, however, it doesn't seem to go all the way.
Here's a long running function:
I'm expecting way more sub-calls so that I may understand why is it running so long.
Here's how that function looks like:
function updateIfNeeded() {
switch (state) {
case 'NO_REQUEST':
throw new Error('Unexpected draw callback.\n' + 'Please report this to <https://github.com/elm-lang/virtual-dom/issues>.');
case 'PENDING_REQUEST':
rAF(updateIfNeeded);
state = 'EXTRA_REQUEST';
var nextNode = view(nextModel);
var patches = diff(currNode, nextNode);
domNode = applyPatches(domNode, currNode, patches, eventNode);
currNode = nextNode;
return;
case 'EXTRA_REQUEST':
state = 'NO_REQUEST';
return;
}
}
It's part of the elm-runtime.
Now, while it is possible that this function might not call any other functions, it would not explain why it's running for so long.
Where is the button for my complete flamechart :}
The Performance panel has a Disable JavaScript Samples checkbox in the capture settings menu.
When this checkbox is enabled, the timeline only shows the yellow placeholders to differentiate between script execution time and, layout, paint and composite activity.
Notice how the Cog/Settings Icon is red when the checkbox is enabled.
When the checkbox is not checked, the timeline shows the flame chart. When all the capture options are in their default state, the Cog/Settings Icon is blue while the menu is open and grey when the menu is closed/collaped.
Unfortunately it is not possible to be certain that this was the exact issue that you encountered, as the shared screenshot doesn't depict the capture settings.
Hopefully this knowledge proves valuable should you encounter the same behaviour in the future.
Related
I am designing a GUI using C, Glade, and Gtk.
I have some signals configured in glade to update the labels of various widgets, mainly GtkButton and GtkLabel. The overall functionality is that when a certain radio button is clicked, all button and labels change in response (language selection).
I am using the function gtk_label_set_label(...) in the widgets _draw() function and it works as expected (text changes, g_print occurs (once)).
gboolean on_lblMyLabel_draw(GtkLabel *label, gpointer *user_data) {
gtk_label_set_label(label, "custom text");
g_print("%s\n", "custom text");
return FALSE;
}
However, when I attempt the same from a button,
gboolean on_btnMyButton_draw(GtkButton *button, gpointer *user_data) {
gtk_button_set_label(button, "custom text");
g_print("%s\n", "custom text");
return FALSE;
}
The text does not update, but dissappears, and the g_print() statement prints forever (as if the draw is recursively calling itself).
Funnily, if I move the button code from _draw to _click, it works as expected, however, I need the GUI to redraw itself, so updating on click is impractical.
Is there a way, using _draw() to prevent this?
Is there a better way to do this?
thx!
Is there a way, using _draw() to prevent this?
No, and you shouldn’t be using the draw signal for this either. It has an entirely different purpose, and will be called each time a widgets redraws itself. That’s also the reason why your button is going into an infinite recursion: you changed its label so it figures it needs to be redrawn; that redraw leads to your callback being called, which again changes the label, etc etc
Is there a better way to do this?
Yes, and you mention it yourself already: make sure you do the logic of changing the widgets in the appropriate place (for example, on a click event), and let the GTK widgets take care of redrawing themselves.
Unless you’re doing something very exotic (like not running an event loop, which you automatically get with GtkApplication), this will all work fine.
I'm using AutoIt to try and make a little hotkey application to work with Windows Journal so I can quickly select different colors.
It seems I'm very close and yet very far to getting the desired result. I've used the AutoIt tool to find the CommandID of the toolbar and the ID of the colors. Here is my code:
ControlCommand("[CLASS:JournalApp]","",113,"SendCommandID", 40178)
My problem is that the color will not be selected. It will be selected to the degree that the color will have the "selection" brackets around it, but the color that I draw with will still be the last color I've selected.
So I tried messing around and found that this code:
ControlCommand("[CLASS:JournalApp]","",113,"Check","")
It will indeed select the color, but it will only select the light blue color. I don't know why, but that is the color that is always being selected. I have not found a way to combine the selecting ability of "SendCommandID" with the checking ability of "Check"
Also, it is a ToolbarWin32 Control.
I figured it out myself.
Here's what I've learned:
ControlCommand("[CLASS:JournalApp]","",113,"Check","")
Has a serious weakness in that it there is no way to specify which button will be checked. At first it seemed to be random, but after a while of playing around I noticed that it did it at a specific coordinate relative to the client window. Why? I have no idea. But at least it's not random.
ControlCommand("[CLASS:JournalApp]","",113,"SendCommandID", 40178)
Has a weakness in that, while on the surface it appears to have successfully clicked the button. The button's function is not actually executed. For my specific circumstance, the color of the pen did not change after I used this, though it appeared to click on the button.
Here's my solution(s):
I looked around and found that AutoIt has a library specifically for dealing with ToolBarWin32 Classes. This is the library from GuiToolbar.au3. With this I found that I was able to do a few things. One, was that I could send clicks to the buttons and change the state of the buttons even. I found that changing the state of the buttons did nothing in relation to triggering an event and the clicking worked, but it had the weakness that it caused the mouse to flinch. This did not work because my pen was near my tablet as that has priority of mouse movement. So I had to raise my pen away from my tablet in order to use the hotkeys--not very convenient. Here was my code for that solution:
if WinActive("[CLASS:JournalApp]") Then
WinActivate("[CLASS:ToolbarWindow32; INSTANCE:2]", "")
$cmdId = "401"&$hotKeys[$key-1+$shift]
If $cmdId < 40172 or $cmdId > 40188 Then
Return
EndIf
$hWnd = ControlGetHandle("[CLASS:JournalApp]", "", 113)
_GUICtrlToolbar_ClickButton($hWnd, $cmdId)
EndIf
What I found after was that AutoIt's native ControlClick() was a lot more useful in that it didn't cause the mouse to flinch whatsoever. It triggered the mouseclick event directly. So that in combination with a nice command from the toolbar library made for a much cleaner solution. Here it is:
if WinActive("[CLASS:JournalApp]") Then
WinActivate("[CLASS:ToolbarWindow32; INSTANCE:2]", "")
$cmdId = "401"&$hotKeys[$key-1+$shift]
If $cmdId < 40172 or $cmdId > 40188 Then
Return
EndIf
ConsoleWrite($hotKeys[$key-1])
$hWnd = ControlGetHandle("[CLASS:JournalApp]", "", 113)
;get the coords of the button and control send a click
local $btnCoords= _GUICtrlToolbar_GetButtonRect($hWnd, $cmdId)
ControlClick("[CLASS:JournalApp]", "", "[CLASS:ToolbarWindow32; INSTANCE:2]","left",1,$btnCoords[0]+2,$btnCoords[1]+2)
EndIf
My question is simple: WHEN (on what event?) can I be sure that a control has fully loaded and has its states and templates also?
Why am I asking:
I'm trying to restore the state of my own WP7 control after recovering from tombstone. This control looks like a calendar in a weekly view. In this calendar you can select many items displayed as colored Rectangles.
If I select any of them, and then go to tombstone and come back to the page, it seems like my control forgot which Rectangles were selected. In fact, it did NOT forget the data itself, but the Rectangles forgot their selected state.
After recovering from tombstone, I try to select the Rectangles by setting their VisualState to "Selected" (which works in any other scenario). I found out, that it fails, because VisualStateManager can't find the "Selected" state.
I know this is tricky, because when coming back from tombstone the controls do not build exactly as in any "normal" case. (for example Bindings and Templates do not apply in the same order) But up until now I could always trust, that when FrameworkElement.Loaded fired, I had my controls ready. Now it seems like VisualState is not. (I tried to set the state from Loaded event handler, but results are the same, VisualStateManager.GoToState returns with false.)
What more can I do?
This is a tricky one! I have also experienced issues where UI events fire before the UI itself is fully constructed, see this blog post for an example. My general approach to this is to handle the LayoutUpdated event, which fires each time the visual tree is updated. You will find that this event fires multiple times, both before and after the Loaded event.
When the Layoutupdated event fires, you can check whether the visual state change has worked, if so, no longer handle the event. If not, keep trying!
Within your loaded event, try the following:
// try to set the state
if (VisualStateManager.GoToState(myControl, "myState") == false)
{
// if failed, wait for the next LayoutUpdated event
EventHandler updateHandler = null;
updateHandler = (s, e2) =>
{
if (VisualStateManager.GoToState(myControl, "myState") == false)
{
myControl.LayoutUpdated -= updateHandler;
}
};
myControl.LayoutUpdated += updateHandler;
}
i'm still having a problem clicking a popup button on an auction site,that appears only if u won an auction. This popup seems tbe a problem. Ive managed to get help partially in Need to click a bid button with Grease monkey script, i'm able to get the bid buttons clicked, but the popu is stil a problem. The xpath for the popup is:
.//*[#id='ctl00_mainContentPlaceholder_Button3']
And the script i'm using currently is:
// ==UserScript==
// #name click popup try1.3
// #include http://www.trada.net/*
// ==/UserScript==
// ctl00_mainContentPlaceholder_Button3
function PopClick ()
{var PopBtn1=document.getElementById("ctl00_mainContentPlaceholder_Button3");
alert('try1');
PopBtn1.click (1);
alert('finished');
};
PopClick();
But the problem seems to be that the script doesnt stay active on the page waiting for the popup, I think if i can get it to "wait" for the popup to appear, it should work. I'm very new to GM, so excuse if there is simple errors. I had exelent help from people like Brock sofar, who is showing me the ropes. Slowly but surely I'm gettin the hang of it. Remove the alerts, i just used them to see if it executing.
The simplest solution would be to run this function, say every second, thus "waiting" for the popup to appear:
setInterval(PopClick, 1000);
It is also better to rewrite PopClick to check if the element is there, before calling click, like this:
function PopClick () {
var PopBtn1=document.getElementById("ctl00_mainContentPlaceholder_Button3");
if(PopBtn1) {
PopBtn1.click ();
// It is also makes sense to clear interval here. see docs for setInterval/clearInterval please :)
}
};
May be this will help you with the freezing issue.
The more proper way, however, would be to set up MutationEvent listener. Since you are using Firefox, it should work OK:
function click_if_popup(evt) {
if(evt.target.hasAttribute('id') && evt.target.getAttrubute('id') =="ctl00_mainContentPlaceholder_Button3")
evt.target.click();
}
document.addEventListener('DOMNodeInsertedIntoDocument', click_if_popup);
Sorry, I didn't test any of this code: I wanted just to give you the general idea of where to dig.
I have a page with possibly several content-editable iframes (editors).
Now I would like to use my custom Firefox extension to do the following:
Setting the cursor to the end (or last HTML element) of the editor the cursor actually is in.
I found many solutions to get the cursor's position, but I need one to set it.
Any suggestions?
XPCOM likely includes such functionality as part of the testing rig. Mochitest at least is capable of this (again, probably though XPCOM).
On the other hand, when a user is on the system this a generally a gross violation of user interaction practices. Be sure you have a good justification for doing it. It may seem convenient but what if they're doing something else whilst using your addon? I usually have various apps open at once, Fx extensions are only part of that. I don't want it taking control of my mouse, EVER.
Is there something wrong with setting the focus? At least that only forces the user's hand at a window level.
It also suspect it make it quite difficult to get past AMO review. You'd have to justify why it was necessary to invoke such low-level functionality. If you interact with a window, for example, the window might be able to affect the input of your functions which in turn control the mouse... and then a random web site has access to the user's window!
Found the solution to my problem myself. This code myself will set the Cursor position to the last Paragraph of my editor:
var frame = window.content.document.getElementsByTagName('iframe')[2];
var win = frame.contentWindow;
var editingSession = Components.classes["#mozilla.org/editor/editingsession;1"].createInstance(Components.interfaces.nsIEditingSession);
var editor = editingSession.getEditorForWindow(win);
selection = window.getSelection();
var body = frame.contentDocument.body;
text = frame.contentDocument.createTextNode(".");
body.lastChild.appendChild(text); // add textnode to be selected
var range = editor.document.createRange();
range.setStartBefore(text);
range.setEndAfter(text);
editor.selection.removeAllRanges();
editor.selection.addRange(range);
body.lastChild.removeChild(text); // remove Child