[Request] Testing on a Mac

chrisglasier

ok, this is weird.

if I don't add the alerts the code doesn't get run. looks like some synchronization problems from HTML world to SketchUp world.

need to debug this further.

Sorry to have lumbered you with this.

The thing works on my PC with or without the alerts (but I expect you would have guessed that).

Do you think I should try and get the Sketchup team to look at it. CMD showed an interest in cgScenes so maybe he (or she) could help. Please let me know what you think.

And thanks again of course.

Chris

Jim

Unless I misunderstand the problem, SketchUp always uses Internet Explorer no matter what other browser is installed.

thomthom

@jim said:

Unless I misunderstand the problem, SketchUp always uses Internet Explorer no matter what other browser is installed.

On PC, yes. But on Mac it uses the Safari engine.
So when you're on PC developing a plugin - with no Mac to test on you can try your code on Safari to do some basic testing.
If you're on Mac it's worse though as IE hasn't been available on Mac since IE5 - and even then it was a completely different browser than the PC version.

thomthom

If this Mac testing isn't too urgent I could have a look at it. But I don't have much spare time as I'm working on an architectural project besides work which is due on ~2 weeks. +pluss my involvement in other projects and various beta testings.
But I do have a Mac Mini at home that I've used to debug ruby plugins.

AdamB

[First off, thanks to TBD for forwarding the link to me..]

I never did really understand what it is you're trying to do, Chris, but the example TBD posted can be made to work by editing line 16 of sceneTest.rb to be:

cmd = "receiveSceneCount('#{count}')"

Guess they're not handling the (admittedly spurious) terminator correctly.

Adam

chrisglasier

Thanks all!

It is pretty late here, and I have only the energy to copy post my reply to Jim's PM ...

Thanks Jim, this Mac business is very troublesome. I found the following in the API documentation:

@unknownuser said:

Finally, there is a difference in the way that the Mac boots up SketchUp that you should be cautious about: there is no Sketchup.active_model when the Ruby scripts are first loaded. So if your script is making changes to the active_model at load time, it will not work on the Mac. The answer? Ensure code that references the active model is part of a UI event handler, responding to the user selecting a tool or a menu item. You can also use an AppObserver to get a callback whenever a new model is opened, at which point it's safe to talk to the active_model.

I looked at the AppObserver class, but as I have no experience of observers and I will have limited time with the Mac tomorrow, I thought if someone could help with the observer mentioned, I might just have a chance.

I'll post this in the topic as well.

My regards

Chris

tbd

that code will never work race condition - JS code will run before the Ruby thing so you will get wrong values. that's why it works only when alert commands are inserted (which is not a solution imho)

you need to refactor the code and take that in account (or push Google to open their implementation of JS bridge from DC)

Jim

http://forums.sketchucation.com/viewtopic.php?f=315&t=19831&st=0&sk=t&sd=a&start=45#p165464

@unknownuser said:

On the WebDialog topic...

We wrote a somewhat generic bridge between JS and Ruby that we used for Dynamic Components. I learned a lot from that and would have some thoughts on how to build a new one that's truly generic and easy to use. That's almost worthy of its own big thread.

One note: I think it would be work for the JS code side of a bridge to be stored in the ruby and then dumped into the WebDialog via execute_script, so JS code side of things would call something like this:

Code: Select all
window.location.href='skp:initialize_bridge';

...and then from there would have access to whatever JS methods we implement.

It doesn't sound like Google wants to open their code, but Scott says he is willing to help design a better bridge.

chrisglasier

Thanks for all your help. In the end with the constraints of machinery and time, I decided to resort to the good old 1995 solution:

It worked for windows



function start(){

if(navigator.appName == 'Microsoft Internet Explorer') browserStart()

else{

confirmPanel = document.getElementById("confirm")

confirmPanel.innerHTML = "START"  }   }



function receiveSceneCount(x){
count = x    }



function browserStart(){


//TO FIND IF ANY SCENES EXIST

....


function confirmDown(){

switch(confirmPanel.innerHTML){

case "START"; browserStart();  break

....

I hope it works but it really doesn't seem right.

Also I think it is wrong to start off a major new excursion into automated interfaces with workarounds! I would really like to find a proper solution. Perhaps the evolution of SKX will help expose it - but I wonder how.

Anyway thanks again for your time; I will let you know what happened tonight tomorrow; I foresee some grief!

Kind regards

Chris

thomthom

hmm... It might be that you try to do changes to the webpage before the DOM is ready. If you do things immediately as the webdialog is loaded you need to wait until all the HTML elements are built.
http://www.javascriptkit.com/dhtmltutors/domready.shtml

chrisglasier

I think I will look at that deeply if the thing fails tonight because everything now seems to work with a PC; and I have loaded the page into Safari and there are no syntax errors (tested with FF as well) just complaints about skp of course. So I don't want to mess with what exists.

It seems to me this is much the same as loading the active_model business. Anyway thanks for the info. I just hope I don't need to re-jig everything!

Cheers

Chris

fredo6

@chrisglasier said:

Could some kind person(s) have a look at this plugin code? I'm told it does not work on a Mac (I only have a PC). I have gone through all the javascript and transformed all IE only syntax. Perhaps a cursory look at the attached files is enough to expose the cause.

Many thanks

Chris

Chris,

You may just be facing a well-known synchronization issue with web dialogs on Mac. Unlike in Windows/IE, the windows.location callback verb called from HTML does NOT block, so that you continue execute JavaScript. Furthermore, most callbacks will be 'lost', as Sketchup uses only one buffer to communicate with Ruby, not a queue.

See that post: http://forums.sketchucation.com/viewtopic.php?f=180&t=13394&p=101428&hilit=+safari#p98793

You have to do the synchronization manually, which, I agree, is not a piece of cake.

Fredo

chrisglasier

@unknownuser said:

You may just be facing a well-known synchronization issue with web dialogs on Mac. Unlike in Windows/IE, the windows.location callback verb called from HTML does NOT block, so that you continue execute JavaScript. Furthermore, most callbacks will be 'lost', as Sketchup uses only one buffer to communicate with Ruby, not a queue.

Thanks for replying Fredo. I am afraid I am quite new to Ruby, generally just using the callbacks to instruct SU to do things - "move component[x]" type of thing. I expected some hassle with Mac/PC compatibility but from the post you quoted it seems it is way beyond my ken, and the problems getting access to a Mac are fraught with frustration (as I predicted). I don't even know if the current version doesn't work but I suspect it won't.

@unknownuser said:

You have to do the synchronization manually, which, I agree, is not a piece of cake.

Does this mean I have to make chains of javascript/ruby functions, viz

JS01 call ruby01; ruby01 call JS 02; JS02 call ruby 02 ... until the play is complete.

I have done something like this before with setInterval functions that control a number of different objects with different timing and paths, so if I have interpreted what you say correctly, I feel I would be still in with a chance (British stoicism and all that).

But there is still the problem, common on the forum I now know, of finding Mac owners willing to test. If anyone reading this will volunteer, I promise a mention in dispatches. Here, opportunistically, is the current "best chance" set of files.

Includes start option for Mac

Incidentally, Fredo, I did try to press for resolution of this well-known synchronization issue again in this reply to CMD - over optimistically perhaps.

My thanks and regards

Chris

chrisglasier

@thomthom said:

If this Mac testing isn't too urgent I could have a look at it. But I don't have much spare time as I'm working on an architectural project besides work which is due on ~2 weeks. +pluss my involvement in other projects and various beta testings.
But I do have a Mac Mini at home that I've used to debug ruby plugins.

That's kind of you to offer. Considering the obstacles bought up here and in the SFX Bridge topic, I think time becomes subservient to finding some solution (even if temporary). Perhaps it could be lodged in the repository for the benefit of anyone else attempting such similar lunacy with web dialogs and Macs.

Two sets of files are attached. Any original files and directories should be deleted. For the Mac version, I have:

... introduced a START control to deal with the fact there is no Sketchup.active_model when the Ruby scripts are first loaded. This is not a tool or a menu item but I think the effect should be the same.

... rewritten the code so that no window.location call is followed by any javascript statements.

... opened the MAC files in FF and Safari to see if either reported syntax errors.

Opportunistically, I included the PC version which includes the makings of a javascript call - initiated by user action in SU's display and selection observer - as described here. A solution to this too might be a good addition to the SFX web dialog section.

Whilst there is hope I am very willing to rejig things, provided I have guidance, so please let me know what you think.

Double tak*

Chris

Norwegian for thank you if I remember correctly from my time working with Snohetta

cgScenes R1.034(MAC).zip

cgScenes R1.035(PC).zip

thomthom

Close!
Tak = Roof
Takk = Thanks

OT (sorry but you got me curious) what work did you do with Snøhetta?

chrisglasier

@thomthom said:

Close!
Tak = Roof
Takk = Thanks

OT (sorry but you got me curious) what work did you do with Snøhetta?

Alexandria Library (design time schedule and monitoring); also co-authored Atomic Architecture with Craig Dykers, one of the founders ... a long time ago!

Chris

Edit added AA link - now Google is working

TIG

Can some one with a Mac tell me what they get returned when they use this snippet in the Ruby Console...

ENV["TEMP"]

Is it 'nil' or a 'path_to_the_temp_folder' ?
Also with

ENV["TMP"]

ditto ...
Also a Mac whole ENV listing, got thus...

ENV.sort

would be helpful.

tbd

ENV["TEMP"]
=> nil

ENV["TMP"]
=>nil

ENV.sort
=> [["COMMAND_MODE", "legacy"], ["DISPLAY", "/tmp/launch-jGxAIU/;0"], ["HOME", "/Users/TBD"], ["LOGNAME", "TBD"], ["PATH", "/usr/bin;/bin;/usr/sbin;/sbin"], ["PYTHONPATH", "/Applications/Quick Search Box.app/Contents/Frameworks/Vermilion.framework/Resources"], ["SHELL", "/bin/bash"], ["SSH_AUTH_SOCK", "/tmp/launch-4hMGhW/Listeners"], ["TMPDIR", "/var/folders/oo/oofmaDN7HfGByvhOdkcyck+++TI/-Tmp-/"], ["USER", "TBD"], ["__CF_USER_TEXT_ENCODING", "0x1F5;0;0"]]

TIG

Thanks TBD, so your temp folder is called "TMPDIR"