Fix multitouch handling to fix onscreen keyboard handling #298

Eeems · 2023-03-27T17:36:33Z

No description provided.

Eeems · 2023-03-27T17:43:03Z

@X-Ryl669 would you be able to test this? One thing to be aware of, is that this has some breaking changes from the stable 2.5 that would make going back not work quite right. A bunch of settings have been moved to the shared settings file. You may want to back up the ~/.config/Eeems folder before installing and running the code from this branch.

X-Ryl669 · 2023-04-04T13:18:22Z

Ok, I've tested this PR (it was a journey by itself), and it still doesn't work.
BTW, the new tarnish.postinst script uses "is-enabled" and "how-to-enable" function in bash script that aren't defined and it fails to install with opkg.

Instead of jumping to the full launcher page, I now get a icon roll on the bottom of the screen containing only Xochitl.

Also, the battery level in the launcher is not the same as the battery level reported by xochitl (71% vs 68% if it's of any use)
I'd say that the bug is harder to reproduce, but it does happen anyway if I insist.

Eeems · 2023-04-04T14:00:34Z

Ok, I've tested this PR (it was a journey by itself), and it still doesn't work.
BTW, the new tarnish.postinst script uses "is-enabled" and "how-to-enable" function in bash script that aren't defined and it fails to install with opkg.

Oh, I guess that might be due to switching to toltecmk, so installib is no longer available. I'll have to source it manually. I wonder why nobody else has reported that when testing PRs for me.

Instead of jumping to the full launcher page, I now get a icon roll on the bottom of the screen containing only Xochitl.

This is the application switcher, you can get to it by swiping up from the bottom of the screen.

Also, the battery level in the launcher is not the same as the battery level reported by xochitl (71% vs 68% if it's of any use)

Xochitl reports the battery as being 0% when there is still around 10% available. I'm just reporting the raw percentage that sysfs outputs.

I'd say that the bug is harder to reproduce, but it does happen anyway if I insist.

It being harder to reproduce is at least better. Could you get some logs with DEBUG=1 when you reproduce it to see what is happening?

X-Ryl669 · 2023-04-04T14:46:45Z

Please refer to the issue page, I've posted more logs and a patch that fixed the issue for me.

X-Ryl669 · 2023-04-04T14:52:05Z

Oh, I guess that might be due to switching to toltecmk, so installib is no longer available

That's what I did, I've followed the README to create a package and ended up with plenty of ipk files I had to install.

But it failed, so I did a opkg update and an opkg upgrade. And then it failed in the upgrade, it upgraded wget before openssl so it couldn't call wget anymore (missing dependency on libssl version 3, wget refused to start). So it couldn't download openssl and libatomic either. I had to download them on my PC, extract them to get the 2 libs that were missing and install them manually in /opt/lib so wget could be launched and opkg could finish its work.

Yet, when trying to install tarnish, the postinst script fails so I guess there are missing dependencies like you said.

Eeems · 2023-04-04T16:03:36Z

Oh, I guess that might be due to switching to toltecmk, so installib is no longer available

That's what I did, I've followed the README to create a package and ended up with plenty of ipk files I had to install.

Instead of the build artifacts on the PR? You can get to them from the Checks tab > Build oxide

But it failed, so I did a opkg update and an opkg upgrade. And then it failed in the upgrade, it upgraded wget before openssl so it couldn't call wget anymore (missing dependency on libssl version 3, wget refused to start). So it couldn't download openssl and libatomic either. I had to download them on my PC, extract them to get the 2 libs that were missing and install them manually in /opt/lib so wget could be launched and opkg could finish its work.

Entware has an issue with the order that wget-ssl and libopenssl get installed, you can fix it fully with the following steps:

opkg remove wget-ssl
wget http://toltec-dev.org/thirdparty/bin/wget-v1.21.1
chmod +x wget-v1.21.1
mv wget-v1.21.1 /opt/bin/wget
opkg update
opkg upgrade
opkg install --force-reinstall wget-ssl

Yet, when trying to install tarnish, the postinst script fails so I guess there are missing dependencies like you said.

Could you check again with the build artifacts?

X-Ryl669 · 2023-04-04T16:34:05Z

applications/system-service/systemapi.h

        }
        auto touch = touches.first();
-        if(swipeDirection != None){
+        if(swipeDirection != None || touch->x == NULL_TOUCH_COORD || touch->y == NULL_TOUCH_COORD){


I've tried this at first, but this fails (some event are missed).

If the event was complete, then the code would generate a DOWN event.
The next report then would generate a MOVE event.

Typically, what happens is that an incomplete touch is received for slot 0 (for example, missing X).
If you don't ignore it in the SYN_REPORT handler, then a DOWN event is generated (it goes to the pressed list).
Then when the complete event is received, a MOVE event is generated while it should be a DOWN event.
So for small sampling, where you'd get DOWN, MOVE, UP from the driver, it's actually treated as DOWN, MOVE, MOVE, UP, with the first DOWN having a invalid position so it'll never be treated as a swipe with this patch.

The result is that when you swipe from the bottom of the screen, it doesn't trigger Oxide's launcher anymore, the log are full of "Swipe cancelled" (well with my previous patches, it was ;-)

Gotcha, I'll just need to fabricate the touchDown call when the X or Y was NULL_TOUCH_COORD and the next event is a move that fills it in.

Didn't my (last) patch worked ?
It's simply patching inside the SYN_REPORT case to ignore it, so everything else works as expected.
It's not dealing with lists so it shouldn't require fixing the malloc stuff size either.

While your patch may work, I'd rather use the modified and existing flags that I implemented the way I intended for them to be used.

Okay, I only set existing if it's a valid event, which will force touchDown handling instead of skipping the touchMove.

X-Ryl669 · 2023-04-04T16:47:34Z

Build artifact doesn't work either (tested before the last commit).

Eeems · 2023-04-04T17:00:34Z

Build artifact doesn't work either (tested before the last commit).

:( Alright, I'll have to fix the build to add install-lib somehow.

Eeems · 2023-04-07T23:35:49Z

@X-Ryl669 have you had a chance to test the latest build?

Eeems · 2023-04-21T22:42:09Z

@X-Ryl669 poke?

X-Ryl669 · 2023-04-22T16:04:13Z

Sorry Nathaniel, I forgot to test it. I'll fetch it on Monday and let you know.

X-Ryl669 · 2023-04-24T16:48:48Z

Ok, I've removed my packages and installed yours. I wasn't able to reproduce the issue, so it's a clear success!

Eeems · 2023-04-24T16:50:25Z

Awesome! Time to merge the PR!

Fix #296

35b4507

Eeems added the bug Something isn't working label Mar 27, 2023

Eeems added this to the v2.6 milestone Mar 27, 2023

Eeems added 2 commits March 27, 2023 12:52

Add code to handle skipping SYN_DROPPED events

2ac5b8c

Fix version reporting to sentry

08b34c4

Eeems force-pushed the issue/296 branch from bbd7d2a to 08b34c4 Compare April 3, 2023 04:58

Eeems linked an issue Apr 3, 2023 that may be closed by this pull request

Using onscreen keyboard can trigger swipe gestures #296

Closed

Handle invalid touch events

c3eebf7

X-Ryl669 reviewed Apr 4, 2023

View reviewed changes

Add install-lib methods, don't set existing if invalid touch event

f38c09a

Eeems merged commit 80eb537 into master Apr 24, 2023

Eeems deleted the issue/296 branch April 24, 2023 16:54

Fix multitouch handling to fix onscreen keyboard handling #298

Fix multitouch handling to fix onscreen keyboard handling #298

Uh oh!

Conversation

Eeems commented Mar 27, 2023

Uh oh!

Eeems commented Mar 27, 2023

Uh oh!

X-Ryl669 commented Apr 4, 2023 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Eeems commented Apr 4, 2023

Uh oh!

X-Ryl669 commented Apr 4, 2023

Uh oh!

X-Ryl669 commented Apr 4, 2023

Uh oh!

Eeems commented Apr 4, 2023

Uh oh!

X-Ryl669 Apr 4, 2023

Choose a reason for hiding this comment

Uh oh!

X-Ryl669 Apr 4, 2023 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Eeems Apr 4, 2023

Choose a reason for hiding this comment

Uh oh!

X-Ryl669 Apr 4, 2023 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Eeems Apr 4, 2023

Choose a reason for hiding this comment

Uh oh!

Eeems Apr 4, 2023

Choose a reason for hiding this comment

Uh oh!

X-Ryl669 commented Apr 4, 2023

Uh oh!

Eeems commented Apr 4, 2023

Uh oh!

Eeems commented Apr 7, 2023

Uh oh!

Eeems commented Apr 21, 2023

Uh oh!

X-Ryl669 commented Apr 22, 2023

Uh oh!

X-Ryl669 commented Apr 24, 2023

Uh oh!

Eeems commented Apr 24, 2023

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

X-Ryl669 commented Apr 4, 2023 •

edited

Loading

X-Ryl669 Apr 4, 2023 •

edited

Loading

X-Ryl669 Apr 4, 2023 •

edited

Loading