Bug 152083

This is a bug in the nvidia driver which doesn't take care about
the BIOS ordering of primary and secondary display devices. I'm pretty
sure you are using the binary only nvidia driver ?

No, this was during installation using the recognition routines in Sax2 and the OSS nv driver.

Hardware:
Athlon XP2100+ in VIA chipset board, not sure on manufacturer, think MSI.
Aoelus FX5900XT graphics card
VGA port Iiyama AS4316UTC
DVI port Hyundai ImageQuest Q17

If it was during the dual-head configuration after installing the nVidia drivers, I wouldn't have been so worried, but this is in the initial configuration, where the OSS driver cannot handle dual head and the Installer doesn't allow you to add the second monitor.

What is confusing is that Sax2/the OSS driver recognise the monitor on the VGA controller, but then use the one on DVI for primary output (the one that gets the mouse pointer).

I'm sorry I'm clueless what is the problem ? The nv doesn't support
dualhead as you already told me... why is it a problem then ? The nvidia
binary driver has that bug interchanging the understanding of primary/secondary
between the BIOS and the nvidia driver (this cannot be fixed because binary)

??

The problem is the way the nv driver/Sax2 detects the monitor if you have a dual-head set-up during installation. I don't disconnect the second monitor from the system during install... I just wait until the initial YOU update and check the nVidia driver, then configure the dual-head set-up when the desktop first comes up.(During install and using the nv driver the two displays are mirrored.)

If somebody does this when they are using two monitors of mis-matched types, if the Sax2 configures for the bigger specification monitor and uses the lower specification monitor, the lower specification monitor will be blanked out (hopefully), but that will be the screen with the mouse pointer.

Basically, if Sax2 detects two monitors, it should make sure it recommends a specification suitable for the one it will actually use...

The behavior you mentioned is not done by SaX2. As you already said
the nv driver is not capable of doing any dualhead. Therefore sax2 cannot
and will not provide a configuration for it. The fact that you see two
mirrored screens even with the nv driver is hardware related. The hardware
itself does that. If you switch to the text console you will see the console
mirrored as well.

The hardware also will take care about the issue if the connected device
is not able to display a mode. In that case you will not get a signal until
a driver with the correct configuration accessing it.

I think we are talking at cross purposes here... Let's forget about dual-head for the moment...

On the device configuration screen of the set-up routine, the first entry is for graphic card and monitor. In the case where there are two monitors connected to the system, the device driver and the install routine correctly detect *one* of the monitors, on my machine, the monitor connected to the VGA port. All well and good.

The Installation routine suggests a monitor driver - in my case, it correctly deduces the model number of the Iiyama monitor connected to the VGA port. All well and good.

The machine then restarts and boots into X. At this point, the secondary display, the one *not* recognised during the install process, becomes the active display, with the mouse pointer. This is wrong, if the set-up has identified the first monitor, shouldn't it use that monitor as the main monitor?

Now, the problem would be if I had a 1280x1024 TFT on the primary connector and a 1024x768 TFT on the secondary connector. At this point, the 15" would shut down because it is being asked to display an image it is not capable of (best case scenario, worst case the monitor is old and doesn't have any safeties built in). To the user, the 15" doesn't display anything, as expected, they chose 1280x1024 because that is the monitor that the install routine detected, *but*, although the 1280 display shows the desktop, it does not show the mouse pointer because the system (be it X, nv or the hardware) is sending the mouse pointer *only* to the monitor that has shut itself down!

So the problem is, the initial set-up is recognising one display, but X itself is using the other display. If the system is giving a monitor make and model number and giving a suggestion for its configuration (1280x1024x24) during installation, the system should stick to using that monitor as its primary display (i.e. the one that it sends the mouse pointer to).

If the OSS drivers are incapable of detecting the second monitor, even though that is the monitor it is using as the primary display, I think there should be a warning in the release documentation that says that if the user has two monitors attached, (s)he should turn one of them off during initial installation, until the proprietary drivers have been installed and they can set-up proper dual-head  functionality.