Gooey GUI - a peculiar way to crash FS!

[trogluddite]

Hi all,

While I was fiddling around trying to reproduce Spogg's strange Ruby GUI shut-down, my little test rig uncovered something else rather bizarre - or rather, crashtastic! This is on FS 3.0.6 in Windows 10, and, for brave souls, here's a demo schematic (note, it is safe to open as posted, I'll warn you in the following instructions when to put your crash-helmet on!!)

Spoggs Ruby GUI problem.fsm

(895 Bytes) Downloaded 816 times

The only module contains a Ruby counter of GUI redraws - separated into those that are caused by the trigger button input ("Triggered"), and all others ("Workspace"). First, turn it on by setting the 'Enable' boolean to true - then click around the workspace, move the module, etc. you'll see how often it gets redrawn without needing a trigger to the input - even for mouse actions nowhere near the module.

We can also see that Navigator strip previews are drawn separately - if you make the Navigator really big so that you can read the text on the module, you'll notice that it's redraw count is often one more than in the workspace.

Now, and this is very important, make sure the module is NOT selected!!! And then, connect the "Debug" text box to the "Debug" output of the module. Now, WITHOUT CLICKING ON THE MODULE, do some more mousey stuff. Not a lot different than before, except that the count in the text box does count the Navigator redraw (and a bit of other stuff I needed for debugging.)

CRASH-HELMET TIME!!!

Select the module, and while watching the counts, have a play with the icon strip at the bottom of the module; try to rename it, for example...

...and then, if it works like it does here, force quit FS and dismiss the "Report this to Microsoft" window!

I guess it's some kind of feedback loop with the redraws, but I haven't a clue why - none of the other behaviour hints that this would be a problem at all. And why only the module icon strip?

[tulamide]

Most of what you experienced is already known (and fixed in FS4).

This includes triggering redraws through the editime and standard inputs not forgetting their initial class (for example if you switch the first string input to be a view, it shows the "V" but is still a string class, others can even be nil).

Another of these annoying things is that an infinite redraw loop is created if you call "output" from within the draw method in a module with outputs triggered by Ruby. For some bizarre reason a larger area than just the output icon is redrawn when an output is triggered, which reaches the view area of the module which leads to a redraw of the view, which leads to triggering an output, which leads to ... I think you got the picture!

Luckily these are fixed in FS4. However, the crash isn't. So if you could make a schematic that's focused on the crash alone with a short comment how to trigger the crash, I would share it with MyCo, the current developer of FS4.

[trogluddite]

Thanks, tulamide; good to know that it's something that's been tackled - and that my "feedback" hunch wasn't too far off. Nice to hear what MyCo's up to as well - I'd figured that he was helping out with 64-bit, but not quite how big his role is, and I've not seem him here since I got back. It's a shame not to have his amazing coding chops around here so much, but he's just the right guy to be bringing FS up to date!

Spogg's GUI crash itself, I'm playing with still, but I've only managed to reproduce it a couple of times, and only while building rather than explicitly running anything, so it's tricky to automate a test as yet. If I get anywhere, I'll put word out (which reminds me; I should demo that weird scheduled "input" call thing that I described recently.)

[tulamide]

Nice to hear what MyCo's up to as well - I'd figured that he was helping out with 64-bit, but not quite how big his role is, and I've not seem him here since I got back. It's a shame not to have his amazing coding chops around here so much, but he's just the right guy to be bringing FS up to date!

I'm so happy that he uses his coding skills for FS itself now, and I can live with his absence here on the forums for this reason. It's mind-boggling what he managed to do already, because he did it all with a codebase that's like 12 or more years old! And he's such a constant, hard worker. Several hours everyday in addition to his day job. And he does it just for one reason: to help Malc getting his feet back on the ground. He's not paid or something!

And there's so many new features already. 64-bit, Ruby 2.5+, very basic -but at least existing- svg support, a complete overhaul of the toolbox and the preset management, uncountable improvements to DSP and esp. Assembler, a vast array of new prims, some new Ruby methods, but first and foremost bug fixing over bug fixing. Even in alpha state, it is already more stable than any other prior version. So much that one person already uses it in commercial production!

Yes, MyCo getting on-board was the best thing that could happen to Flowstone!

[RJHollins]

WOW .... that is the best News we've heard about FS ... ever.

[trogluddite]

To the above, re: Myco and FS4, I can only add; !

@tulamide
What's the best way to contact MyCo with bug-reports etc. - bearing in mind that I'm a Luddite who doesn't do Facebooky social media thingies! Will he get them if I e-mail support@... ?

[wlangfor@uoguelph.ca]

Very useful to know, thanks Trog

[trogluddite]

Here's a quick demo of the other bug which I mentioned earlier. Note that the situation where this happens is something very rarely used (and 'scheduleMethod' is an easy and reliable workaround) - so don't lose too much sleep over it!

scheduled_input_bug.fsm

(2.15 KiB) Downloaded 805 times

If anyone could give this a quick test on v3.0.6 and above before I add to MyCo's workload, that would be cool.

When it opens, click the "RUN" button, and in the text box you should see lots of Midi objects flashing by in the String box; and it happily runs (FS 3.0.6) for as long as I've had the patience to watch. "STOP" the test run, then change the "Keep References" boolean to false. "RUN" again, watching the display, and after a short while, you should see non-Midi values flashing past, followed shortly after by a crash.

[tulamide]

Will check it soon, trog. Meanwhile, you received a pm

[Spogg]

When I click Run I don’t see those numbers flashing by. It just sits there with nowt happening.
I tried your sequence in any case but still nowt. I added an ASIO out and enabled it but still dead as a dead thing.

Windows 7 32 bit, core i7.

Sorry!

Cheers

Spogg