In T11627#219720, @meven wrote:

In T11627#219718, @ahartmetz wrote:

You may want to look at the DBus part of this with dfer-monitor, a tool I have written. It is especially useful for looking at latency problems.
https://cgit.kde.org/dferry.git/

It seriously lacks documentation, not even a README.

I have been using Bustle for dbus monitoring.

You may want to look at the DBus part of this with dfer-monitor, a tool I have written. It is especially useful for looking at latency problems.
https://cgit.kde.org/dferry.git/

@fvogt Yeah thanks, that made more sense than you spoon-feeding me the information for the commit message ;)

So what do we do here? The Qt bug has been fixed but still I'm not sure what should be done about the environment variables.

In D26185#584249, @davidedmundson wrote:

Thanks for following up!

I'll open a Qt bug and see how it goes.

Can you link it please.

In D26185#583991, @anthonyfieroni wrote:

Can forcing bigger font DPI make things better?

It's probably okay to still say that scaling is enabled "in general" but for scaling <2, and why not, maybe even greater, hinting should not be inhibited. The screen is still made of pixels. Qt is not going to be able to throw out all the hinting and subpixel code like Apple did, because it still needs to be supported for non-Retina screens anyway, for a long time. I'll open a Qt bug and see how it goes.

I think I see the problem. You pay for the possibility to have scaling with unhinted fonts. I am pretty sure that that isn't an acceptable way of handling it. Most Linux desktop fonts are intended to be used with some hinting (slight hinting usually?). QHighDpiScaling::isActive() returns true even when the scale factor is 1.0 as in my case. Looks like a Qt bug to me.

Just to make this extra clear, I have nothing against auto-scaling. There is probably a KCM to set it however I want if I don't like the default. My problem is with blurry and in some cases ("-=") disfigured font rendering. That should never happen regardless of physical and logical DPI (though no subpixel rendering and maybe no hinting might be fine at >= 2.0 scaling). With this patch, it looks like with Qt 5.13 on not very high DPI screens, i.e. it looks nice. I don't own a really high DPI screen so far.

In D26185#582095, @anthonyfieroni wrote:

Test with qputenv(QT_ENABLE_HIGHDPI_SCALING, "1") it's especially for 5.14+

Compiled against Qt 5.14 before this change:

This fixes session restore for me.

@kwin

This change makes kwin_x11 crash on startup. Please fix kwin or degrade this to a warning.

OK, I've tested it, the fix that this amends still works as well.

@sandsmark , it would be good to retest if your fix still works after this. IIRC QApplication does not lazily initialize the event loop though.

There are many more plain qDebug()s in this module, but this one seems to produce the most noise...

Change base revision

Change second index check to >0

Maybe it is better to return an empty string in this case but I don't know. Otherwise I might thave pushed this directly. In the case of my missing file, the returned string was the name of the folder <file_db_data>/00/, which generates an error message that the file it tried to open is actually a folder...