Bug 30002

Xconfigurator IDs chip properly, as does the /proc filesystem.

Really like appearance of Anaconda; can we refactor Xconfigurator to call that
beastie instead?

Can you post the lspci output?

Can you post the output of ddcprobe?  That way, we can tell exactly what the
installer thinks the video card is.

Also, the output of lscpi -nv would be very helpful.

Created attachment 11380 [details]
lspci -vvx output

Created attachment 11381 [details]
ddcprobe's output

By looking at the relavent section of lspci...

01:00.0 VGA compatible controller: Silicon Integrated Systems [SiS] 6306 3D-AGP
(rev a3) (prog-if 00 [VGA])
	Subsystem: Silicon Integrated Systems [SiS] SiS530,620 GUI

It appears that the card thinks that it is a 620.

Also, can you post the output of 'lscpi -nv'.  Using the 'n' option will display
the pci id numbers.  Then we can look them up in the pcitable and see what the
pci ids match up to.

Created attachment 11383 [details]
lspci -nv output

I agree it's amusing. On the other hand, it says it's a 6306 on the line above.
Who do you trust? :-)

It's an on-the-board video chip -- I'd tell you the motherboard id, but I'm
*responding* on the offending machine right now.

Ok, here's the relevant entries in the pci-table:

0x1039  0x6300  "Card:SiS 630"  "Silicon Integrated Systems [SiS]|SiS630 GUI
Accelerator+3D"
0x1039  0x6306  "Card:SiS 620"  "Silicon Integrated Systems [SiS]|6306 3D-AGP"


Since the card's pci id is 1039:6306, this maps to the SiS 620.  This may or may
not be correct...I don't know why Xconfigurator would do it differently, though.

Bill, any thoughts on what to do here with the pcitable?

Not really; it uses the same driver entry under XFree4, so it
really shouldn't make any difference.

What are the symptoms of how it fails/what do you change to make it
work?

Uh... The text on the screen is nearly unreadable. As in, X server unusable :-)
You fix it by overriding anacondra's guess and tell it which video chip it
really has

Can you post the working X config and the non-working X config?

I'm curious, because according to what I see here, changing the mapping
would still leave the X configurations *identical*.

Created attachment 11480 [details]
The XF86Config built by anaconda if I override to SiS630

Created attachment 11481 [details]
XF86Config if I let anaconda configure things as a SiS620

Created attachment 11482 [details]
anaconda builds the XF86Config-4 for SiS620 but not for 630. Odd...

Quit asking questions like that! I had to do a pristine install to make sure I
got a good XF86config
file (and yeah, I *did* push the resolution up a little on the '630)

So, in the working case, are you using XFree86 3 or XFree86 4?

XFree86 4.02. At least, I did NOTHING to tell it I wanted 3.6 (is that even an
option?)

I'm beginning to think that the XF86Config-4 file is vulnerable to being
removed. It's a different topic, and I don't know the causailty. I'm doing a
pristine install, and will make backup files. As soon as I figure THAT one out,
will share.

It's a long story, but has to do with "respawning too fast" messages appearing.
I'll report it separately as soon as I have a story to tell..

We (Red Hat) really need to fix this before next release.

Okay. Story is figured out. To wit: anaconda thinks it's an SiS 620 but I know
better and override it's guess to SiS 630. Anaconda installs XFree86 3.6. Had I
left it alone it installs XFree86 4.0.2

1. 4.x installed for 620 is appalling(sp) If I could do a screenshot I would,
but I'd be lucky to see the outline of a box in this form, and maybe horz lines
for text. Forget mouse movement being predicable
2. Xconfigurator does not re-establish /etc/X11/X -> your_server all of the
time. So the other day
it wrote a 3.6 XF86config, deleted the XF86config-4, but kept 'X' pointing to
XFree (4.0)

Xconfigurator and the XF86.org's configuration tools ought to be given heave-ho
in favor of an
anaconda entry point that only runs the X configuration portion of the tool.
Except! Said tool must 
move your XF86Config out of the way and use a stock VGA one while it runs (and
then restore your's back), because obviously if you can't read the screen with
the current setup, bringing up anaconda just means another unreadable X tool.

If anaconda "knows" that it's going to use 3.6, it'd be nice if it said so on
the GUI somewhere. Or if you could choose between servers (where the given
chipset is equivalently well-supported by both), maybe a preference button would
be cool.

The cards database has been updated to use XFree 4 for the SiS 630, so this
should resolve the issue about using different severs for the differnt cards
when they both can use the XFree 4 servers.  However, if the XFree 4 drivers are
messed up, then this only makes the problem worse.  Preston, should we revert
the SiS 630 and 620 to use XFree 3 by default?

Updated the Cards database to use XFree 3 for SiS 620 and SiS 630.  This should
fix the problem.