atitvout does not work with the current version of X in unstable

[Josh Triplett <josh.trip@verizon.net>]

atitvout does not work with the current version of X in unstable.  See
the following IRC log for details:

<JoshTriplett> Overfiend: Short version: atitvout works fine under
stable (plus a 4.2.1 backport), but has some serious problems under
current unstable.
<JoshTriplett> Overfiend: Same version of atitvout.
<JoshTriplett> Overfiend: And same kernel, because I'm using my stable
chroot.
<joshk>        atitvout seems to have nothin to do with X
<Overfiend>    I've never heard of atitvout.
<joshk>        because i used it to enable tv output on the tty
<JoshTriplett> Overfiend: It's a program to turn on the ATI tv-out, by
doing unspeakable things with /dev/mem, VM86, VBE, and the BIOS. :)
<Overfiend>    maybe atitvout pokes registers that X didn't used to,
until I updated the driver.
<joshk>        that and it sucks, but i haven't bothered reporting bugs
<Overfiend>    JoshT: without knowing more, there's my theory
<JoshTriplett> joshk: Works fine under console too.
<JoshTriplett> Overfiend: I see.
<Overfiend>    is atitvout designed to play nicely with the X server, or
did it merely get along accidentally before?
<JoshTriplett> Overfiend: I have no idea. I don't think it really cares
what is running.
<Overfiend>    that's probably not good
<joshk>        i think it got along accidentally
<Overfiend>    the X server drivers expect absolutely nothing else to
even imagine dreaming of touching the card's registers while it has the VT
<JoshTriplett> Overfiend: The symptom is that the TV-out has strange
brightness/contrast problems, and so does the LCD when I switch back.
But if I then use xrandr to switch video modes, it goes back to normal.
<Overfiend>    as a famous general once said, "it's a court-martial
offense to even imagine to dream of blowing that bridge up"
<Overfiend>    JoshTriplett: sounds very much like something in support
of my theory
<JoshTriplett> Overfiend: Yeah.
<Overfiend>    JoshTriplett: is atitvout stateful?
<JoshTriplett> Overfiend: 1) Is there anything I could do to confirm
that theory? 2) Is there anything I can do to otherwise diagnose or fix
the problem?
<Overfiend>    or does it just poke the card and then exit?
<JoshTriplett> Overfiend: It pokes the card and then exits.
<Overfiend>    JoshTriplett: that's bad.
<JoshTriplett> Overfiend: For example, "atitvout -f t" turns on tv out.
<Overfiend>    I think this is probably a design flaw in atitvout.
<JoshTriplett> Overfiend: It uses LRMI, if that means anything to you.
<Overfiend>    I would mail debian-x and atitvout@packages.d.o and ask
what should be done about this
<Overfiend>    no wonder I haven't heard of it
<Overfiend>    we don't have LRMI on powerpc
<JoshTriplett> Overfiend: Or BIOS, for that matter. :)
<JoshTriplett> Overfiend: It's definitely x86 only.
<Overfiend>    JoshT: this may not be a fixable bug -- it sounds like
atitvout isn't designed to respect anything else
<JoshTriplett> Overfiend: Entirely reverse-engineered, since ATI won't
give TV-out specs.
<Overfiend>    right, because Macrovision is the best company in the world
<JoshTriplett> Overfiend: Exactly. :(
<JoshTriplett> Overfiend: My impression is that they can't, not that
they won't.
<JoshTriplett> Overfiend: s/they/ATI/
<Overfiend>    it's starting to sounds like a ubergeek's tool, not
something mere mortals should be playing with
<JoshTriplett> Overfiend: The internals are definitely deep magic.
<JoshTriplett> Overfiend: But I know no other way to turn on TV-out for
ATI cards.
<JoshTriplett> Overfiend: If you do, I'm all ears.
<Overfiend>    well, I mean that it's a proof of concept that should
spur further development in the X driver or GATOS or whatever
<Overfiend>    JoshTriplett: I don't :(
<JoshTriplett> Overfiend: So you think that perhaps the functionality of
atitvout could be added to X in a stateful way?
<JoshTriplett> Overfiend: And in the meantime, you just want me to mail
debian-x@l.d.o and atitvout@p.qa.d.o ?
<JoshTriplett> Overfiend: Or should I file a bug, so it isn't forgotten?
<Overfiend>    JoshTriplett: 1) possibly; 2) yes
<JoshTriplett> Overfiend: (On atitvout, presumably, from what you've said.)
<Overfiend>    JoshTriplett: what you're asking for is a wishlist
request. things acting weird after atitvout has poked the hardware is
pretty much unfixable
<JoshTriplett> Overfiend: True, but I would think it would be an
important bug on atitvout.
<Overfiend>    vorlon: yes
<JoshTriplett> Overfiend: (Or grave, if it doesn't work for anyone else.)
<Overfiend>    JoshTriplett: if it's not designed to be stateful, maybe not.
<Overfiend>    JoshTriplett: maybe just a big fat warning in the manpage
<Overfiend>    "This tool is a kludge to get around the extreme evil of
the ATI and Macrovision companies."
<Overfiend>    "As such, it does kludgey things. Don't switch back to
your X VT without turning TV out back off. You have been warned."
<JoshTriplett> Overfiend: Actually, that isn't the problem.
<JoshTriplett> Overfiend: It won't work in X at all.
<JoshTriplett> Overfiend: Which renders it almost useless for me.
<Overfiend>    ah.
<joshk>        JoshTriplett: it still lets you put 'bb' on a tty
<Overfiend>    JoshTriplett: I would urge the maintainer to document
these caveats in the manpage or even the package description, then.
<joshk>        and tha'ts pretty cool ;)
<joshk>        er
<JoshTriplett> Overfiend: As in, if I start in X, in a resolution it can
handle, and then run it, it doesn't work.
<joshk>        bb from tty on tv
<JoshTriplett> Overfiend: That's a pretty big limitation.
<Overfiend>    JoshTriplett: please do mail the info from this
conversation, though. I want this in the mailing list archives so Google
will find it.

[I am not subscribed to debian-x; please CC me on replies.
Mail-Followup-To is set accordingly.]

- Josh Triplett

Attachment: signature.asc
Description: OpenPGP digital signature