SketchUp and OSX Mavericks....
-
Yes
I now have the typo - passed from another source...
That must be Ericโs Web-guyโs typo... he added that 'remote' code to an earlier test setup I wrote... however, it still works fine on all OSs [except Mavericks] with AND without that typo ???
I confirm that the corrected 'allow_actions_from_host' code does now work on Mavericks as well as all other OSs... most puzzling...
BUT this begs the questions... as to why very similar code fails in the SketchThis toolset, when its 'allow_actions_from_host' IS identical to the url that its ruby specifies...
??? -
Thanks John - I think that is the typo in question.
As to why it works in all other OS's except for Mavericks.....that I don't know.
-
Oh, TIG already responded. Yes, this is an odd one.
-
I started this discussion so I might as well put in my two cents. My experience with Mavericks so far has been generally positive. In regard to SketchUp I installed Make while keeping SU8 on hand for now. SU8 doesn't seem to run any different for me (as far as I can tell) on Mavericks than it did with Mountain Lion. However, Make does seem to run somewhat better (as far as I can tell) with Mavericks than SU8 does with Mountain Lion.
One odd thing, though. My SketchUcation Store plugin doesn't connect anymore in SU8. I just get a message saying V2 is available, but when I try to click for the update I just get an error message. I can still install extensions the old way by downloading them manually and using the Extensions Manager found in SU Preferences. I've even tried installing the V2 of SketchUcation Store through the Extensions Manager and I still get the error message afterwards when I try to use the plugin. The SketchUcation Store plugin works fine in Make.
-
@chris fullmer said:
Oh, TIG already responded. Yes, this is an odd one.
Chris,
It is easy to scan the WebDialog framework and confirm that this error message emerges from it, not from a Mavericks system library or framework. But without reverse engineering the WebDialog code (which for ethical reasons I won't attempt), it is impossible for us outsiders to know what WebDialog is trying to do when it throws this error. I have built and think that I understand the sample custom protocol handler that Apple provides, but it does not trap any issues with the URL. Without disclosing some proprietary info, would it be possible for the Trimble engineers to tell us what logic test fails when WebDialog throws this error? That info would help to track down what about Mavericks and certain URLs is peculiar to the problem.
As a different matter, it seems like this topic has gotten quite technical - maybe it should be moved or restarted in the Developers forum?
Thanks
Steve -
Yeah, the developer's forum is a good place for it I think.
Still looking into it on our end, nothing new to add to the discussion yet.
Chris
-
It's already in the Devs' forum
-
I just had my mind blown
-
For what it's worth, I just noticed that the same typo was present in both the TIG.remote and TIG.local methods, but only the TIG.remote failed due to the typo. Does allow_actions_from_host() look at previously loaded pages to see whether that host is used by any of them? How does the SketchThis ruby obtain the host URL it passes to this method?
-
-
And of course the local version didn't need the' allow' anyway - it was a web-guy/Eric tinkering with my original ruby
Correcting the typo in the remote version does not change things anyway - the remote still fails with Mavericks... but the local version succeeds [with the unnecessary 'allow' method removed, on with the typo corrected or the typo intact]BUT the weird thing is that if you try the [corrected OR typo-ed] tests on a pre-Maverick MAC or a PC, then both the local and remote versions perform just fine, as if the 'allow' and/or typo is ignored and the 'allow' is 'an irrelevance'
-
Actually, it does just seem to be exactly that, allow_actions_from_host.
I had noticed with V6 that it was ignored, so I never implemented it.
But in my remote WebDialog, which is failing, I simply add an allow_actions_from_host, and suddenly now it's working again.--J
-
I'll jump in now, I concur that it all appears to be related to the dlg.allow_actions_from_host method. I think its possible that Win and OS have not actually been honoring that method. But Mavericks now suddenly does.
So if people are not using it, it will dis-allow certain actions with remote websites and throw that error.
If you do use it (which you really should now), make sure you have the syntax correct. You should be giving it a domain, not a url. example:
dlg.allow_actions_from_host("http://www.sketchup.com")
is not correct, it should bedlg.allow_actions_from_host"www.sketchup.com"
Now I'm curious to know if that is enough to fix all/any of the problems, or are people still seeing other problems?
Chris
-
@tig said:
Correcting the typo in the remote version does not change things anyway - the remote still fails with Mavericks... but the local version succeeds [with the unnecessary 'allow' method removed, on with the typo corrected or the typo intact]
BUT the weird thing is that if you try the [corrected OR typo-ed] tests on a pre-Maverick MAC or a PC, then both the local and remote versions perform just fine, as if the 'allow' and/or typo is ignored and the 'allow' is 'an irrelevance'
OK, this just keeps getting weirder
On my Mavericks Mac, TIG.remote works just fine after the typo is corrected. What the bleep is going on????
-
@slbaumgartner said:
OK, this just keeps getting weirder
On my Mavericks Mac, TIG.remote works just fine after the typo is corrected. What the bleep is going on????
That is what we're seeing here as well Steve. From what we're seeing this all boils down to either having not included the allow_actions_from_host either not being implemented at all, or having been implemented with bad syntax (a url vs a domain string), or a typo.
It appears no other OS, other than Mavericks, actually pays any attention to this method. That is what my quick testing is showing. I'm very open to hear about exceptions to these findings or issues that might be different.
Chris
-
@chris fullmer said:
@slbaumgartner said:
OK, this just keeps getting weirder
On my Mavericks Mac, TIG.remote works just fine after the typo is corrected. What the bleep is going on????
That is what we're seeing here as well Steve. From what we're seeing this all boils down to either having not included the allow_actions_from_host either not being implemented at all, or having been implemented with bad syntax (a url vs a domain string), or a typo.
It appears no other OS, other than Mavericks, actually pays any attention to this method. That is what my quick testing is showing. I'm very open to hear about exceptions to these findings or issues that might be different.
Chris
Whew, at least I might not be losing my mind
I read the documentation for this method multiple of times and I am not at all sure I understand it
What sort of "actions" could take place on a "host remote from the host where the WebDialog exists"? That seems to say that an action callback could be routed to a remote host? How? The skp: protocol handler is part of the WebDialog code running on the local host...very confused...
-
After playing around with this a bit more, I may have answered my own question. On my Mavericks Mac, the skp:callback from TIG.remote fails with the now familiar error message if the call to allow_actions_from_host is omitted or invoked with a bad argument when the WebDialog is set up by Ruby (which is what Chris said before, I just didn't understand).
So, the concern isn't about executing an action on a remote host, it is about preventing skp: callbacks when the WebDialog contents were loaded from a remote host and might be malicious. Restricting the callbacks to pages from only trusted domains registered with this method then makes sense.
Which, of course, doesn't answer the question of why this method is ignored on most OS's and suddenly started working in Mavericks! But that will have to be answered by the Trimble SketchUp engineers, since the rest of us have no insight as to how this part of WebDialog is implemented.
As a note to the Trimble engineers: it would have been a lot easier for us outsiders to track this down if the error message was clearer. The message reports the full URL of the callback that it doesn't like (which is why we see the skp:callback@ part), and calls it an "Invalid URL" whereas it is actually a prohibited callback from a page that wasn't vetted by the Ruby code.
-
So to recap...
If a webdialog uses a remote url and it uses
allow_actions_from_host(string)
and that 'string' is somehow incorrectly formed, then it is ignored by all OSs which continue to operate correctly with its callbacks - except for Mavericks which always fails !
But with a correctly formed 'string' that is a 'domain' Maverick succeeds too.
Mavericks will fail if the dialog is using a remote url BUT it does not include aallow_actions_from_host(string)
, or of course if it is used and the 'string' is not a proper 'domain'.In the case of SketchThis's dialog that host 'string' [given by the web-developer] included an initial
http://
- making it a 'url' rather than a 'domain', with all OSs it had previously worked OK anyway, but now it failed when used with the new Mavericks...
This 'string' has now been corrected in the SketchThis code, it is now a 'domain' without an initialhttp://
, and it now works correctly with the SketchThis dialog callbacks.In contrast with the SketchUcation's Plugin Store dialog that 'string' was already correctly formed as a 'domain', and therefore it worked on ALL OSs, including Mavericks.
So anyone experiencing issues like this with remote url dialogs Mavericks-only should check that their dialog includes the
allow_actions_from_host(string)
AND that it has a good 'string' that is a 'domain' and not a 'url' [or a typo!]...
This good-practice will also future-proof their code against any IE updates that might replicate the somewhat excessive care that is now taken by Maverick's version of Safari... -
I have experienced a few bugs. I'm not entirely if they are related to Mavericks, but they seem to have appeared after the update.
- My tool palettes ( Layers, Style, Scenes, etc ) don't "stick" together anymore. When I click on them to enlarge, they just dissociate from the others, which is pretty annoying.
- I have experienced a few other visual bugs while using Mission Control (with the gestures) ; palettes don't come back as they should, or just slower performances.
-
@tig said:
So to recap...
...
So anyone experiencing issues like this with remote url dialogs Mavericks-only should check that their dialog includes theallow_actions_from_host(string)
AND that it has a good 'string' that is a 'domain' and not a 'url' [or a typo!]... ...Hum... will this also correct the "Java Script disabled" problem ?
Advertisement