GuideMe (TeaseMe v2.0) - Current Build 0.4.4

[RemiHiyama]

That reminds me. There is another issue I've noticed with hotkeys. If you are on page of a tease that has buttons with hotkeys and open the debug window to jump to another page (like while debugging or testing), if you start typing in the dropdown box to enter a page name, if that page name uses any of the hotkeys, the hotkeys trigger that button, even though technically the Guideme window no longer has focus, the debug window does. You have to jump to some safe page that doesn't have hotkeys, and whose name doesn't use any of the characters used as hotkeys on the active page.

I see what you mean - if you run GuideMe with java instead of javaw in the batch file so you get a command window, you can see all the listeners firing no matter what window you have open. There's some code that appears intended to prevent that from happening, but it's obviously not working and I don't really know why it's being done the way it is.

This one might be an EroticDevelopment problem.

[philo]

That reminds me. There is another issue I've noticed with hotkeys. If you are on page of a tease that has buttons with hotkeys and open the debug window to jump to another page (like while debugging or testing), if you start typing in the dropdown box to enter a page name, if that page name uses any of the hotkeys, the hotkeys trigger that button, even though technically the Guideme window no longer has focus, the debug window does. You have to jump to some safe page that doesn't have hotkeys, and whose name doesn't use any of the characters used as hotkeys on the active page.

I see what you mean - if you run GuideMe with java instead of javaw in the batch file so you get a command window, you can see all the listeners firing no matter what window you have open. There's some code that appears intended to prevent that from happening, but it's obviously not working and I don't really know why it's being done the way it is.

This one might be an EroticDevelopment problem.

It was done like that, because I hadn't done professional coding in ten years and was learning Java to get back into programming again. A lot of Guideme was written when my skills were very rusty, in a language I hadn't used before, so a lot of it was just the first thing I got to work

[PlayfulGuy]

Spoiler: show

That reminds me. There is another issue I've noticed with hotkeys. If you are on page of a tease that has buttons with hotkeys and open the debug window to jump to another page (like while debugging or testing), if you start typing in the dropdown box to enter a page name, if that page name uses any of the hotkeys, the hotkeys trigger that button, even though technically the Guideme window no longer has focus, the debug window does. You have to jump to some safe page that doesn't have hotkeys, and whose name doesn't use any of the characters used as hotkeys on the active page.

I see what you mean - if you run GuideMe with java instead of javaw in the batch file so you get a command window, you can see all the listeners firing no matter what window you have open. There's some code that appears intended to prevent that from happening, but it's obviously not working and I don't really know why it's being done the way it is.

This one might be an EroticDevelopment problem.

It was done like that, because I hadn't done professional coding in ten years and was learning Java to get back into programming again. A lot of Guideme was written when my skills were very rusty, in a language I hadn't used before, so a lot of it was just the first thing I got to work

Having learned python to write my new downloader I can totally relate to that. Many times I've had to go back and revisit code as I learned, or now I look at code I wrote a year ago and with the benefit of what I've learned I can see way better ways of doing what I did back then. And they don't always get updated

And it's good to see you're still stopping by and commenting now and then philo. Hope life is treating you well.

PG

[RemiHiyama]

It was done like that, because I hadn't done professional coding in ten years and was learning Java to get back into programming again. A lot of Guideme was written when my skills were very rusty, in a language I hadn't used before, so a lot of it was just the first thing I got to work

Ah.

Guess the question is why it stopped working, since it seemed fine in 4.4, and I didn't think anything had touched that code... Well, a different approach is probably indicated anyway, since if you've had library window open and then closed it (ie used it at all), it'll start spitting out error messages for every keypress as it tries to get the text of a widget that's been disposed of. Presumably the same thing happens if you close the debug window.

[PlayfulGuy]

It was done like that, because I hadn't done professional coding in ten years and was learning Java to get back into programming again. A lot of Guideme was written when my skills were very rusty, in a language I hadn't used before, so a lot of it was just the first thing I got to work

Ah.

Guess the question is why it stopped working, since it seemed fine in 4.4, and I didn't think anything had touched that code... Well, a different approach is probably indicated anyway, since if you've had library window open and then closed it (ie used it at all), it'll start spitting out error messages for every keypress as it tries to get the text of a widget that's been disposed of. Presumably the same thing happens if you close the debug window.

This has been an issue in all versions of Guideme that I've used. I think it's always been like this, I just never got around to mentioning it since it really only pops up when you're trying to debug an issue with a tease that happens to use hotkeys.

PG

[RemiHiyama]

This has been an issue in all versions of Guideme that I've used. I think it's always been like this, I just never got around to mentioning it since it really only pops up when you're trying to debug an issue with a tease that happens to use hotkeys.

I really don't know what to make of that, since it's definitely not happening in my copy of 4.4. Maybe something else about our installs is causing different behavior?

[EroticDevelopment]

I do plan on removing these as discussed. I don't believe any of that was included in the 0.4.5 release I published, but if I missed something let me know.

I did some digging and it looks like that never got put back into master on my fork, so there's nothing code-side for you to do; I can just pull that out and submit the backend stuff at a later date, since I think it's still pretty solid and that repeating and/or fractional timers are useful things to have.

But there's still a line in the release notes at https://github.com/EroticDevelopment/Gu ... tag/v0.4.5 about repeating timers that should be removed.

Good catch, removed.

... And also the fix for https://github.com/EroticDevelopment/GuideMe/issues/33 seems to have gotten lost somewhere in there. (It's this commit.)

After some digging, it looks like this change was likely accidentally reverted in the Revert Timer Enhancements commit. I looked through that one but didn't notice this was a part of that revert commit. I'll make sure this is put back in the next release.

I've been spending some time looking at this but so far haven't made any major breakthroughs. I kind of know what's causing it, just trying to figure out how to fix it any why. This issue and the video playing in its own window are largely the same issue I believe. Somewhere it's not attaching the video surface correctly.

Will it actually run for you? I've had crashes on launch with my attempted fixes, but I'm not certain the problem isn't something missing/misconfigured in the VM I'm using for testing the linux build.

Kind of not really? I'm building it on Windows (using mvn's "-P linux64" to set the profile) and then attaching the debugger remotely from my IDE on Windows. It works but isn't ideal. Since I'm running these VMs in VirtualBox it's not as bad. I can share the build folder to the VM and port-forward the debugging port. I initially ran into a crash-on-launch bug that drove me nuts for about an hour. I finally rolled all the way back to 4.4 to try to find the bug. 4.4 worked, so I re-built 4.5 and magically it worked now too. I have no idea what changed. I completely removed the build directories to no avail before all this, so I really can't figure out what was cached causing the issue. I have to think somehow mvn was hanging on to something and causing the issue.

[EroticDevelopment]

Hmmm, I believe the term I'd use is "overzealous" - it makes sense some of the time, but it's doing it all of the time whether it makes sense or not.

In any event, adding some new options to control this behavior is definitely on my list of things to look into. (Well, technically it was already there, since I don't entirely like how it does things either, but you've brought some things to my attention about it that I hadn't noticed before.)

I'm glad to hear that. I always feel like a dick when nitpicking other's work.

I definitely know what you mean having spent time on both sides of this one, but don't feel bad. Feedback is what drives a significant part of development on applications like this. As a developer it's also frustrating when there's a small bug that would take <5min to fix and you find out months later it's been detracting from the application.

Btw I think I found a bug. Disabled buttons can still be triggered by their hotkeys.

Yep, that's a bug. If Remi hasn't resolved this one I'll take a look while I'm looking at some other hotkey issues mentioned above.

[EroticDevelopment]

That reminds me. There is another issue I've noticed with hotkeys. If you are on page of a tease that has buttons with hotkeys and open the debug window to jump to another page (like while debugging or testing), if you start typing in the dropdown box to enter a page name, if that page name uses any of the hotkeys, the hotkeys trigger that button, even though technically the Guideme window no longer has focus, the debug window does. You have to jump to some safe page that doesn't have hotkeys, and whose name doesn't use any of the characters used as hotkeys on the active page.

I see what you mean - if you run GuideMe with java instead of javaw in the batch file so you get a command window, you can see all the listeners firing no matter what window you have open. There's some code that appears intended to prevent that from happening, but it's obviously not working and I don't really know why it's being done the way it is.

This one might be an EroticDevelopment problem.

Yeah, I can take a look at this. There's a lot of event listeners in GuideMe and it's hard to follow at times. I'm guessing some listeners just need to be removed/unbound at times or a variable introduced to return and not perform any actions.

[EroticDevelopment]

That reminds me. There is another issue I've noticed with hotkeys. If you are on page of a tease that has buttons with hotkeys and open the debug window to jump to another page (like while debugging or testing), if you start typing in the dropdown box to enter a page name, if that page name uses any of the hotkeys, the hotkeys trigger that button, even though technically the Guideme window no longer has focus, the debug window does. You have to jump to some safe page that doesn't have hotkeys, and whose name doesn't use any of the characters used as hotkeys on the active page.

I see what you mean - if you run GuideMe with java instead of javaw in the batch file so you get a command window, you can see all the listeners firing no matter what window you have open. There's some code that appears intended to prevent that from happening, but it's obviously not working and I don't really know why it's being done the way it is.

This one might be an EroticDevelopment problem.

It was done like that, because I hadn't done professional coding in ten years and was learning Java to get back into programming again. A lot of Guideme was written when my skills were very rusty, in a language I hadn't used before, so a lot of it was just the first thing I got to work

Hey, to echo what PlayfulGuy said it's great to see you stopping by and I hope all is going well.

Looking back on work you've done while new to a language is always interesting. I have my share of projects that I look back on and wonder why I ever did that and at my day-job role of merge-master would definitely not accept lol.

[PlayfulGuy]

This has been an issue in all versions of Guideme that I've used. I think it's always been like this, I just never got around to mentioning it since it really only pops up when you're trying to debug an issue with a tease that happens to use hotkeys.

I really don't know what to make of that, since it's definitely not happening in my copy of 4.4. Maybe something else about our installs is causing different behavior?

Well that'll teach me to check my facts before I speak! I just did a test in version 4.3, 4.4 and 4.5, and they are all working perfectly today, so you're absolutely right.

I actually moved to a new computer about a week ago and hadn't tried that since then, so it must have been something messed up on my old computer.

I'll make sure I test first next time.

PG

[RemiHiyama]

I looked through that one but didn't notice this was a part of that revert commit. I'll make sure this is put back in the next release.

Yeah, I'm not quite sure how that happened.

As for the rest of the timer stuff, I restored the branch for that on my fork, so I can submit a pull request with the good parts of the timer stuff at some point in the not-too-distant future.

Kind of not really? I'm building it on Windows (using mvn's "-P linux64" to set the profile) and then attaching the debugger remotely from my IDE on Windows. It works but isn't ideal. Since I'm running these VMs in VirtualBox it's not as bad. I can share the build folder to the VM and port-forward the debugging port.

Sounds like you're doing better than I am, I couldn't get host extensions to build so just getting development builds into the vm is a hassle. But I think I'd about reached the limits of my java skill as regarding the linux issues anyway.

Yep, that's a bug. If Remi hasn't resolved this one I'll take a look while I'm looking at some other hotkey issues mentioned above.

It looks to me like just replacing the places that have

Code: Select all

if (event.display.getActiveShell().getText().equals(shell.getText()))

with

Code: Select all

if (event.display.getActiveShell() == shell)

will fix this one. (And it should get rid of the issue with the Library window too, since it won't be trying to call getText on a disposed widget.)

[EroticDevelopment]

It looks to me like just replacing the places that have

Code: Select all

if (event.display.getActiveShell().getText().equals(shell.getText()))

with

Code: Select all

if (event.display.getActiveShell() == shell)

will fix this one. (And it should get rid of the issue with the Library window too, since it won't be trying to call getText on a disposed widget.)

I agree, but something makes me nervous... Checking equivalency against `shell` instead of `shell.getText()` seems easier and more intuitive, so it concerns me that Philo did it this way for a reason or to fix a different bug. When you check equivalency to an object, Java requires it to be the same instance / same heap address. The existing code is looking to see if it's an object of the same class rather than the same instance. I want to make sure that nuance isn't important here before changing it. If this ends up being important we can always just check that `shell` isn't null or disposed.

[RemiHiyama]

I agree, but something makes me nervous... Checking equivalency against `shell` instead of `shell.getText()` seems easier and more intuitive, so it concerns me that Philo did it this way for a reason or to fix a different bug.

More or less how I felt about it, actually.

When you check equivalency to an object, Java requires it to be the same instance / same heap address. The existing code is looking to see if it's an object of the same class rather than the same instance.

Not so; the current code is comparing the window titles. And, uh, it seems to be working now; I'm no longer able to replicate the original bug. Still, comparing the objects seems to be the more correct way to handle the issue, based on what I was able to find other people getting told to do. (See for instance https://stackoverflow.com/questions/376 ... -triggered )

(There are apparently circumstances where getActiveShell can return null, particularly on linux, but my impression is that "not run the hotkey handler" would be the desired behavior in cases anyway.)

I want to make sure that nuance isn't important here before changing it. If this ends up being important we can always just check that `shell` isn't null or disposed.

I don't know about important, but the LibraryShell issue is annoying. If you've opened and closed it repeatedly, you'll end up with multiple pages of error messages from the leftover listeners. We should probably be removing those...

[EroticDevelopment]

When you check equivalency to an object, Java requires it to be the same instance / same heap address. The existing code is looking to see if it's an object of the same class rather than the same instance.

Not so; the current code is comparing the window titles. And, uh, it seems to be working now; I'm no longer able to replicate the original bug. Still, comparing the objects seems to be the more correct way to handle the issue, based on what I was able to find other people getting told to do. (See for instance https://stackoverflow.com/questions/376 ... -triggered )

Reading this back I did a pretty terrible job of explaining what I meant here. You're correct, the current code is comparing window titles. While my statement about the existing code checking if they're the same class definitely doesn't hold for Java as a language, it does within the confines of GuideMe. Each shell class sets a unique title in the function that creates it, and that title is never changed. So, comparing shell titles should behave exactly the same as comparing class names for shells in GuideMe.

I went back and looked at Git history for this line, and it doesn't appear that it was intentionally changed to this for any particular reason. It was changed from something completely different when the Library window was added back in November of 2018 here. So, I think we're safe to change this. Even if you can't reproduce the bug right now, I can see how this could break and it's an easy adjustment.