Re: I'm want to improve hwsetup and hwdata-knoppix
On Tue, Sep 21, 2004 at 02:24:13AM +0300, AKL. Mantas Kriauciunas wrote:
> Hi again,
>
> On Fri, Sep 17, 2004 at 02:13:11PM +0200, Klaus Knopper wrote:
> > On Fri, Sep 17, 2004 at 11:10:28AM +0300, AKL. Mantas Kriauciunas wrote:
> > > - why pcitable from hwdata-knoppix doesn't use standart Redhat hwdata
> > > pcitable format and are there any plans to improve hwsetup to use standart
> > > pcitable format ?
> >
> > Ahem... Could you specify in what way you assume the formats would differ?
> > IMHO, they don't. In fact, I'm just using RedHat's original pcitable and add
> > additional lines, and exchange some that are known to be incorrect by the
> > correct ones.
>
> As I told in my letter - main difference is in pcitable - third column
> "Card: Video_Card_Model" in lots (but not in all) of cases is replaced with
> "Server:XFree86(x)"
Both possibilities are valid and evaluated by hwsetup. I still see no big
difference, sorry.
> > > - are there any plans to use latest hwdata package from Redhat/Fedora or to
> > > merge it with current hwdata-knoppix (current redhat version is 0.130 and
> > > contains lots of updates comparing with latest knoppix version, for
> > > example more than 50 new graphic cards in pcitable) ?
> >
> > As soon as the Debian maintainer of hwdata puts a new package in
> > Debian/unstable.
>
> In Debian unstable (and also in Sarge) newer versions (0.117, 0.118, 0.123)
> are long time ago :)
Can't be more than a few weeks. I got the last update end of august.
> There are several differences between latest hwdata-knoppix and hwdata
> from Redhat/Fedora or Debian, for example in 0.123 version from Debian
> there is this line:
>
> 0x1002 0x5159 "Card:ATI Radeon 7000" "ATI|Radeon RV100 QY [Radeon
> 7000/VE]"
>
> while in hwdata-knoppix info about same device is incorrect:
>
> 0x1002 0x5159 "Server:XFree86(vesa)" "ATI|Radeon QY"
Because the previous entry was reported NOT TO WORK with the radeon or ati
module, I had changed it to "vesa".
> > Usually, the first one is taken in this case. Which other "lots" do you mean?
> > Can you just send a diff -u, please?
>
> I can send diff, but I think there will be more benefit if you start use
> pcitable info from Redhat/Fedora/Debian hwdata package, especially where
> in hwdate-knoppix is used vesa or unknown driver.
No, because there are entries in the "original" pcitable that are KNOWN TO BE
WRONG. I will continue to modify the pcitable in its genuine format, according
to reports of users who have the hardware. Part of the seemingly successful
hardware detection are lots of reports by users who contribute by reporting
their hardware, and corrections to pcitable that work for that hardware.
> I noticed, that in some lines there is 2 pciids in one line (in original
> hwdata pcitable, for example for pciid 0x102b 0x0525). In header of
> pcitable is this info:
>
> # The format is ("%d\t%d\t%s\t"%s"\n", vendid, devid, moduleName)
> # or ("%d\t%d\t%d\t%d\t%s\t"%s"\n", vendid, devid, subvendid, subdevid,
> # moduleName)
>
> I don't know what is subvendid and subdevid, I think we should ask
> redhat/fedora developers if noone from debian-knoppix mailing list knows.
Does not really matter, because hwsetup never uses the subvendid data actively.
Maybe libkudzu does, somewhere.
> > This is apparently not true, if you just have a look at the database.
> > Could you send me the definite lines that cause problems for your hardware,
> > with a patch to use the _definitely correct_ lines instead?
>
> I've looked at the database lots of times. As I already told (see above) -
> there are several lines, where in original pcitable correct driver is
> specified, but un hwdata-knoppix there is vesa :(
Sorry, this is insufficient. I need first-hand information. So, if YOU have a
specific card that WORKS WITH YOUR MODIFIED pcitable LINE, please send me a
patch. If you just SUSPECT that a different pcitable line would be better,
without verificytion, please find some hardware to try first before reporting.
> > Please keep in mind that RedHat's original pcitable may contain entries that
> > require proprietary drivers which are not present in Knoppix, or have been
> > superseded by other lines that have been sent in by users who had problems
> > with the original setup.
>
> AFAIK fedora is true open source project, so pcitable from fedora
> shouldn't containt entries, that require proprietary drivers.
What is your source of that information? Have you checked the fedora XFree86
source versus the one shipped with the RedHat "Enterprise" edition personally,
and compared the differences?
Again, I modify pcitable lines when there are first-hand reports from owners of
specific hardware, or when I can verify those entries by myself, using the
Debian XFree86 server (for graphics cards).
> Also if
> some users had problems with the original setup, then this should be
> documented somewhere, because from me experience original setup was
> correct in more cases, than hwdata-knoppix.
Yes, it should be documented. I would really like to have a webform for
pcitable/Cards/Monitors modifications with detailled bug reports. But so far,
the changelog supplied with the hwdata package, plus the email reports that I
get from owners of hardware, are my only source of information.
If you would like to contribute a website and form for this, I would really
appreciate it. Maybe someone else already has, but I did not find it yet.
> > There are no "drivers". hwsetup only identifies the correct XFree86 module for
> > changing /etc/X11/XF86Config-4 appropriately.
>
> I know this ;)
I wasn't sure if you did. Sometimes I forget. ;-)
> > In which way would you change hwsetup? I may be misunderstanding your mail,
> > but nothing I have read so far would require a change to hwsetup.
> >
> > hwsetup also parses the "Cards" database, if no "Xserver()" entry is present,
> > and takes the matching XF86Config-4 entries from there.
>
> I didn't notice, that hwsetup also parses the "Cards" database, now
> things are more clear - it seems hwsetup takes the matching XF86Config-4
> entries from "Cards" database only if there are matching drivers in
> /usr/X11R6/lib/modules/drivers/ directory.
Yes.
> In my knoppix-based
> distribution this directory doesn't exist at detection time, so only
> entries with "Xserver()" were used correctly.
So, in fact, hwsetup's behavious WAS correct, because it uses only the
"accellerated" entries if the matching modules are found at runtime. And vesa or fbdev as
fallback.
A solution to your problem could be just adding symlinks to all drivers you
have, into the /usr/X11R6/lib/modules/drivers/ directory.
> Solution is to add run
> time option for this.
Well, adding options like "--nocheck" would be a workaround. But not a good
one, in my opinion.
> My friend Alex commented out driver files
> detection (for changes look at
> http://cvs.sourceforge.net/viewcvs.py/morphix/hwsetup-morphix/hwsetup.c?r1=1.1&r2=1.2 ) and now hwsetup takes the matching XF86Config-4 "Cards" database :)
>
> If you accept I can write patch to add option "test for available XFree
> drivers" to hwsetup.
I don't really need that option for Knoppix, which always has all available OS
drivers for XFree86 present, but if nothing else is changed in hwsetup, I would
add your patch (if you could kindly send one) to the repository.
Regards
-Klaus Knopper
Reply to: