[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: Plans for X



On Sun, Oct 22, 2000 at 05:22:11PM -0700, Steve Bowman wrote:
> I had some problems with X 3.3.6 on cvs hurd last night I was going
> to report, then your message arrived so I'll combine that report with
> my response....

Okay, let's see...

> On Mon, Oct 23, 2000 at 12:53:39AM +0200, Marcus Brinkmann wrote:
> > On Fri, Oct 20, 2000 at 08:54:18PM -0700, Steve Bowman wrote:
> > > planning to rework your patches for 4.x,
> > 
> > Ok, I looked into it, and it doesn't seem to hard. There are all old issues,
> > so the 3.3.6 patch is very applicable (unfortunately nobody took the time to
> > find out, and simply rejected it outright), but there are a couple of new
> > issues.
> > 
> > The current[1] status is: Everything compiles, except xload, but that's easy to
> > make work (I have a similar patch for exim somewhere). But the resulting
> > binary doesn't run properly: There are some socket problems (starting twice
> > won't work, you have to reboot. X believes another server is running), and
> > the keyboard handling doesn't work (it gets interrupted system call).
> 
> Hmmm.  I just tried to run 3.3.6 last night on hurd built from cvs[1].

I don't know which CVS you mean, so I don't know which sources you actually
compiled.

> The system completely hangs - at least it appears to at the console.
> I was able to telnet in from elsewhere to see if I could get any useful
> information.  X was already down (i.e., no X-related processes running), I
> tried to restart X remotely and it thought another server was running

I found out that this is related to pflocal. Killing pflocal can help.
I don't know what the real cause is. It seems that X dying fast without
cleanup blocks some socket in pflocal (it should probably free it itself).
But I don't have details, and I am over this now, because my X 4.0 server
doesn't crash anymore *grin*

> It may be that the display or keyboard just need resetting to get the
> console back, but I haven't figured out how to do that and had to reboot.
> Is there a way to reset the display and keyboard after X has crashed
> without rebooting?

No easy way. I am sure you could hack a program to reset everything, but
there doesn't exist one.
 
> This sounds suspiciously similar to the problems you've had with X 4.x.

Actually, only the "it thought another server was running", but yes.

> I'm wondering if there's something going on with the hurd that X is
> triggering.  What's changed since the 20000921 hurd except the tunnel
> changes?

Nothing. I suspect that your build is broken, but I don't know which sources
you actually compiled. No official source or repository I know of carries my
patches.

> Or should I be cross-compiling hurd instead of trying to
> build natively?

I don't think that's helping either. I am pretty sure the Hurd is sober
except some non-critical pflocal weirdness.
 
> I haven't tried restoring the old hurd yet to verify that X works again,
> but I actually did the same thing last weekend and restoring the old
> (0921) hurd cured X. I didn't report it yet because I wanted to rebuild
> and reinstall hurd to make sure the problem wasn't there.  AFAICT,
> it's a clean build except as noted in [1].

Mmmh. If you really can verify that it works with the 0921 hurd, but not
with your new one, I will retract what I said above :)

> > Anyway, as I said, those are attackable. Especially as I got it to compile.
> > 
> > I will follow up with real patches. If somebody has the guts to help out, I
> > can make preliminary patches available soon, but only if you really want to
> > debug things. Otherwise just wait for the packages to become available and
> > stick with XFree 3.3.6 for now.
> 
> I'm game to work on it[2],

The current patch is attached. It is not 100%, but almost. X compiles and
runs, but you have to -allowMouseOpenFail. That's not a problem, though,
because the mouse handler is broken anyway. I also don't get a full display,
but only the bottom half of my screen contains the top of the bitmap. But
keyboard works.

So there is something funky with the display stuff and the mouse handler.
but apart from that, X 4.0 works fine with the patch (as far as I could test
it with half the screen and no mouse).

> in fact, I was going to ask for your 3.3.6
> patches to see what kind of things I should look for and where to look
> in 4.x to patch.

I have something better, the 4.0 patch, see above. All of the upstream
related changes in the 3.3.6 patch are integrated. BTW, this patch is for
the official 401 tar file, not for the Debian source. I don't have the
Debian source here currently, so I can't work from that. Will do it tomorrow
or later this week. But progress is steady, and I am not concerned about the
Debian packaging, it's not hard. What needs work is the display and mouse
stuff, and if you can help fixing that, great. But if you don't have much
time, I'll probably beat you to that because I really want to push this
through.

> Although my comments to Branden were in reference to
> some other boxen I have with an NV1 chipset (and therefore not supported
> by 4.x), my hurd box has a Diamond Viper V770 card (TNT2 chipset) that
> should work with 4.x.

Mmh. Maybe my display trouble is a graphic card issue? I have a Hercules
Terminator, which is Trio64 based. I use it with the vga driver.
I will have to test under Linux, so I don't hunt a bug that isn't there...

> I've looked at the 3.3 source but I haven't looked
> at the 4.x source yet nor have I fooled with Branden's 4.x packages so
> 4.x is new territory for me even if X is not.

It's not as worth as I feared. Much of X we are concerned about stayed the
same, one prominent exception being the input device handling, hence the
problems with the mouse. That part changed drastically.

One final note. Sometimes it is failing to open the kbd device (interrupted
system call or even computer bought the farm error). This is probably a bug
in the kbd translator. Anyway, trying several times usually leads to success
anyway.

My primary goal is to get the upstream source in shape and commit the patch
to the X group and Branden. Then the Debian packaging will be adjusted.

Thanks,
Marcus

-- 
`Rhubarb is no Egyptian god.' Debian http://www.debian.org brinkmd@debian.org
Marcus Brinkmann              GNU    http://www.gnu.org    marcus@gnu.org
Marcus.Brinkmann@ruhr-uni-bochum.de
http://www.marcus-brinkmann.de

Attachment: xfree-401.diff.gz
Description: Binary data


Reply to: