SketchUp and OSX Mavericks....

[Warped9]

I've read enough reviews as well as others stating their SketchUp running fine with Mavericks that I've decided to go for it. I'm downloading now.

Still, while the free download and upgrade is nice I miss the old hard disc install which I felt was much faster.

[Warped9]

I've installed Mavericks and everything seems to work fine, particularly including SketchUp. So far so good.

I do have one glitch unassociated with SketchUp. The upgrade kicked me off my wi-fi network and I can connect using only the guest connection. But when I enter the password to reconnect to my primary network connection it won't go. I've tried the install disc and nothing seems to work. Argh!!!

[timinder]

I have SU 2013 and use it with dual monitors. I have all my menus and info windows opened in the Left screen but now whenever I open SU, the main toolbar pops over to the RH edge of the main (RH) Screen. Anyone else experienced this?

[Warped9]

Mavericks seems to have done something to my display somehow. It's not a problem, but images seem crisper now, including when working in SketchUp. Everything else is as it was before.

[TIG]

Has anyone tried using SketchUp web-dialogs, like SCF Plugin Store, with Mavericks installed ?
I've had a report that callbacks are busted on a similar tool's web-dialog's downloads...
Any Ruby Console errors ?
I suspect that the Mavericks Safari js is not handling the callback as it has before when passing skp:callbackname@urlstring and generating errors...

[aarondietzen]

I recently upgraded to Mavericks and MOST of SketchUp runs fine, though there does seem to be an issue with Safari windows (which are used for some plugins in SketchUp). I have not gotten Ruby errors because of it, but Safari seems to be doing something odd to empty spaces typed into text boxes in dialogs.

Other than that, there is a problem with a serious lag on DisplayLink video adapters.

[bmike]

TIG, seems to load fine. No time to play with downloading... will try later.

[TIG]

What I have just seen reported is that at least in some tools' server-side callbacks, it passes the whole callback command-string as if it were the string-part [after the @] - e.g. the SketchUp callback ' callbackname' is correctly called, BUT it is passed the id & skp:callbackname@http_urlstring_to_load_skp instead of the usual id & http_urlstring_to_load_skp.
Running the callback on PCs or older MACs works fine, but with Mavericks there will be a Ruby Console error that the URL is invalid [and of course it is invalid as a url when it's prefixed with " skp:callbackname@"]

The web-dialogs should still load just fine, but it's the callback part etc that fails when the callback's string gets 'mangled' - breaking linked downloads, linked page opening etc. If any of you test this please do so with the Ruby Console open and post any error messages [or successes!]

[slbaumgartner]

@tig said:

What I have just seen reported is that at least in some tools' server-side callbacks, it passes the whole callback command-string as if it were the string-part [after the @] - e.g. the SketchUp callback ' callbackname' is correctly called, BUT it is passed the id & skp:callbackname@http_urlstring_to_load_skp instead of the usual id & http_urlstring_to_load_skp.
Running the callback on PCs or older MACs works fine, but with Mavericks there will be a Ruby Console error that the URL is invalid [and of course it is invalid as a url when it's prefixed with " skp:callbackname@"]

The web-dialogs should still load just fine, but it's the callback part etc that fails when the callback's string gets 'mangled' - breaking linked downloads, linked page opening etc. If any of you test this please do so with the Ruby Console open and post any error messages [or successes!]

Didn't hit this in any of the tools I normally use. Can you identify specific offenders?

[TIG]

The first report I got was yesterday: it is from the SketchThis toolset and site-page component_load tool...
Their HTML guy compiles the server-side's possible skp download lists into HTML coded buttons using: <a href="skp:callbackname@http_url_path_to_skp" ... etc
On the Ruby-side the callback is fired and it receives the url string to the right of the @.
This works fine using PC and MAC.
It fails with Mavericks - the passed string after the @ seems to be replaced with the whole callback string including the 'skp:callbackname@' part, so the url is then invalid.

I have asked them to look at changing to a simple header defined js function
function callbackname(url){ window.location='skp:callbackname@'+url }
that is in turn the called in the linked buttons instead of hrfe=... as
onClick="callbackname('http_url_path_to_skp');"

It will continue to work in PC/MAC and hopefully in Mavericks too...
I am at a loss to see what Maverick's Safari is doing differently

Because the Plugin Store's web-dialog uses similar 'onClick=' callbacks [rather than the 'href=' way], then I'd like some feedback from Mavericks users - specifically that the AutoInstall download/install works as expected, and the 'More Info' button link also opens OK - this will show that the callbacks are still getting the correct url passed to them...

[sketch3d.de]

@slbaumgartner said:

...but I don't think Apple changed anything that should matter to SU.

Apple has added the OpenGL v4.1 core profile support (recent would be v4.3) in OS X 10.9 Mavericks, which could have an effect on the SU display output.

btw, if you wanna test the OGL tesselation speed of your graphics accelerator you may run the free GpuTest benchmark utility.

Norbert

[slbaumgartner]

@tig said:

The first report I got was yesterday: it is from the SketchThis toolset and site-page component_load tool...
Their HTML guy compiles the server-side's possible skp download lists into HTML coded buttons using: <a href="skp:callbackname@http_url_path_to_skp" ... etc
On the Ruby-side the callback is fired and it receives the url string to the right of the @.
This works fine using PC and MAC.
It fails with Mavericks - the passed string after the @ seems to be replaced with the whole callback string including the 'skp:callbackname@' part, so the url is then invalid.

I have asked them to look at changing to a simple header defined js function
function callbackname(url){ window.location='skp:callbackname@'+url }
that is in turn the called in the linked buttons instead of hrfe=... as
onClick="callbackname('http_url_path_to_skp');"

It will continue to work in PC/MAC and hopefully in Mavericks too...
I am at a loss to see what Maverick's Safari is doing differently

Because the Plugin Store's web-dialog uses similar 'onClick=' callbacks [rather than the 'href=' way], then I'd like some feedback from Mavericks users - specifically that the AutoInstall download/install works as expected, and the 'More Info' button link also opens OK - this will show that the callbacks are still getting the correct url passed to them...

Both Autoinstall and 'more info' worked fine for me.

I've always used the window.location=... technique, not href. I'll look around to see if I can find another example (or contrive one). Do you suppose it is related to link directly from HTML vs window.location setting from javascript? It doesn't make sense to me that this difference would change the way the skp: protocol processes its arguments, but...

[slbaumgartner]

@tig said:

The first report I got was yesterday: it is from the SketchThis toolset and site-page component_load tool...

I just installed the SketchThis toolset and clicked their tool links. They are sometimes very slow (looks like a blank page), but I got to their site. Have they already rewritten this?

[ericschimel]

Nope, we haven't re-written anything. As TIG says, it works fine on the PC, and the Mac pre Mavericks. For some reason it just won't work on Mavericks. TIG developed the Ruby for me, and I have web developers that built the website portion of that, as I said, we haven't changed anything.

I just downloaded the trial of Keyframe animation from the Extension warehouse, and it has the EXACT same issue on Mavericks that my plugin has. Watch the Ruby console and you'll see the same error that my plugin has when you try to hit "Insert into model"

It would appear that this problem isn't just limited to my plugin, so there are potentially a fair number of plugins out there that could be broken.

While I'm not much of a programmer, (that's why I've got TIG!) I am pretty good with testing things and using debug tools, so if there is anything I can help the community out with, I am more than happy to do that.

If it's any help, here's a video of me trying to see if it was a web traffic issue:

https://www.dropbox.com/s/8l4yt6bswy22b80/Mavericks%20SketchThis%20Bug.mov

And here's a screenshot of the ruby console error from my plugin:

http://note.io/1a5H9F5

And here's a screenshot of the same error in Keyframe animation:

http://note.io/1gNdwlo

Again, if there is anything I can do to help out, please let me know.

[slbaumgartner]

@ericschimel said:

And here's a screenshot of the ruby console error from my plugin:

http://note.io/1a5H9F5

And here's a screenshot of the same error in Keyframe animation:

http://note.io/1gNdwlo

Again, if there is anything I can do to help out, please let me know.

Aha, that's the difference! The prior info didn't lead me to understand that the error is on "Insert into model". When I do that, I get the same error. Unfortunately, since the plugin is scrambled ruby I can't actually debug it.

Mavericks includes a new version of the WebKit library (I just checked), so they must have changed/broken something that impacts the way that the skp: protocol is spliced in.

[martychristoffel]

I'm running into palettes that are no longer "sticky" since upgrading to Mavericks. To be more precise, although the palettes do snap to one another, when expanding one such as the Layers palette, it opens underneath the minimized Styles, Scenes, etc. instead of having those palettes remain connected to the Layers pane. When minimizing them down to their title bars, they can also lose their relative association to other palettes.

[slbaumgartner]

@slbaumgartner said:

@ericschimel said:

And here's a screenshot of the ruby console error from my plugin:

http://note.io/1a5H9F5

And here's a screenshot of the same error in Keyframe animation:

http://note.io/1gNdwlo

Again, if there is anything I can do to help out, please let me know.

Aha, that's the difference! The prior info didn't lead me to understand that the error is on "Insert into model". When I do that, I get the same error. Unfortunately, since the plugin is scrambled ruby I can't actually debug it.

Mavericks includes a new version of the WebKit library (I just checked), so they must have changed/broken something that impacts the way that the skp: protocol is spliced in.

There must be something more going on in this Tool than I am privy to. I wrote a little test ruby and html page that uses <a href=skp:callback...> to generate a callback to Ruby, and I get the parameter string as usual, without error, including when it is the exact URL in your page. I do not get the skp:callback@http://... stuff reported by the error message. Also, I found that error message string in the WebDialog code, so I think it is thrown back at Ruby from WebDialog when the Ruby code does something to try to load the skp into the model.

TIG, I think this is back in your court, since the Ruby is scrambled.

[TIG]

I too have written some simple href="skp:callbackname@string" examples, which I thought might break Mavericks.
They didn't.
Then I rewrote the SketchThis code so that it puts into the Ruby Console the string that is received from the 'install' click href=... callback.
Surprisingly it never gets to this point, instead it just presents the "Runtime error: invalid URL 'skp.callbackname@http_url_path.skp'", then stops - never displaying the string as it never gets a valid callback returned.
This is plain weird...

[Aerilius]

Did you try variations on how the webdialog's html is created or loaded?
In the Trimble/Google forum, someone suggested it's due to set_url(); which would make sense if the Maverick's security model has been further restricted.

[TIG]

But my own test-rb+html , the SketchUcation Plugin Store AND SketchThis all use set_url()
Only SketchThis fails with Mavericks.
It uses:
href="skp:callbackname@http://url.skp"
Which was the initial 'suspect', since the SketchUcation Plugin Store is reported as working OK and that uses onClick and js function with the equivalent of:
window.location="skp:callbackname@http://url.skp"

However my tests include various permutations of 'clicking' to callback - i.e. href=... and window.location=... etc - and ALL of those work on Mavericks, and return the passed string correctly, whereas the very similar looking html code href=... for SketchThis fails with the 'invalid URL' message in the Ruby Console ???