Is there a reason to fatal()?

[iUltimateLP]

Is there a reason to fatal()?

[natinusala]

This fatal thing is used to raise "fatal" errors that cannot be recovered. I don't want to just log something and move on as it will easily be overlooked by users.

The right thing would be to use the "crash" dialog, but again what would happen if the crash dialog itself crashes?

Do we have the ability to spawn the OS crash dialog with our own message in Horizon? That might be the best option here. And for Windows / Linux, open a generic crash dialog using the OS API.

[p-sam]

Yes, we have the ability to invoke the error applet, but first it should be checked (I have never used it) that it doesn't send telemetry on its own (or if it used to on older versions), and that it allows to show a custom message, it might be just an error code for example

[iUltimateLP]

Yea, we do have a few options in libnx/error.h, however @p-sam is right, seems like the functions create an error report which is logged to the system, which doesn't seem optimal.

Personally, I dislike the whole Atmosphere crash screen, as it won't give you any info at all, whether it was an exception (and it's desc maybe) or anything else in most cases.

Maybe we can have a lightweight error dialog inside borealis itself - and as a last resort fallback, it'll fatal().

[natinusala]

To be clear I am not talking about the fatal sysmodule but about the "the software has closed" one

[iUltimateLP]

Ah, well that "The software has closed" dialog is fine, I was developing in applet / album mode, and there we just straight up get a Atmosphere crash instead of the dialog (at least as of Atmosphere 0.12, I'm a bit behind).

[natinusala]

This is not normal and should be fixed

[DarkMatterCore]

Crashes in applet mode always trigger the Atmosphère fatal screen. IIRC it's related to the way applets work - I don't think there's an easy fix for that.