QT EGLFS with bogus EDID, force setting mode to unsupported 1024x768
-
Hi,
The question is rather simple. Right now, we are using QT for a medical device (we use yocto with QT and a integrated small computer using an Intel CPU with the integrated GPU).
We want to use EGLFS (no Xorg/Wayland) to simply clone displays. Basically, we clone the display "eDP1" to "HDMI1". The thing is, the display connected to HDMI1 has bad EDID informations (the EDID only reports one mode of the 5 modes supported, so it only sends 1280x1024@60Hz). Except we want to use 1024x768@60Hz. Since device cloning needs the two displays to have the same resolution, when one display is 1280x1024 (the one with the bad EDID) and the other one is 1024x768 (the "embedded display", which has correct EDID), the cloning fails.
I wonder if there would be a way to get around this. Basically "forcing" the 1024x768 mode on the external/bad-edid display (HDMI1) even though it is "theoritically" unsupported according to the EDID (we know it is in fact supported!).
Right now we simply use the variable QT_QPA_EGLFS_KMS_CONFIG, in which we have a JSON that tells QT to use mode "1024x768" with eDP1, and clone eDP1+use mode "1024x768" on HDMI1.
The thing is, from the logs in QT, it seems QT ignores the asked "1024x768" in the JSON because the EDID doesn't report it, thus QT defaults to "1280x1024" (the single and only mode contained in the bogus EDID) and since the two displays don't have the same resolution, we get an error message indicating QT can't clone eDP1 on HDMI1 properly.TL;DR How do I force a mode/resolution not supported (or at least contained) in the EDID? Is there a way to do it in QT C++ code? Elegantly?
-
Hi and welcome to devnet,
Did you try to enable the kms related logging category to see if you get some more information ?
The possibly best way to do what you want is to patch the backend code and implement the "force use" of the selected geometry.
-
Hi, thanks for your answer!
I indeed did enable KMS logging, which is why I came to understand that QT tries to set one display at 1280x1024 and the other one at 1024x768 (which, since device cloning is on, is technically invalid behavior). -
So I maintain that you should modify the backend code to add an enable "ignore the EDID informations" through an environment variable.
