Le Sigh...

[swestrup]

So, the server is apparntly still borked (as is my spelling -- today you're just gonna have to live with it.) The last thing we did before _sps_ left on Sunday night was to install new versions of lots of the software, including X. Now it doesn't start. Or, rather, it sort-of kind-of starts, but not really.

I suspect that there are now two different X implementations or something running at the SAME TIME and its all gone monkey. The login screen is now text only (except for a light-gray mandrake background -- WTFFFFF?) and I can't run anything that wants X. If I say something like startx then it hangs for 10 minutes and finally says something like :0 is already in use, should I try :1?

If I say Yes, and do an Alt-Ctl-F1, I get a normal login screen that sometimes boots me into ICE, and sometimes runs Gnome.

Yeeeesh.

I'm sure this mess would only take 10 minutes to fix for someone who had even the vaguest notion of what the software stack looks like from the hardware up to Gnome, but I don't.

For today, color me defeated.

Comments

[sps.livejournal.com]

PS you want to do a man startx. I forget if it's startx :1 or startx -- :1, but there's definitely a way of telling it what you want from the get-go.

[swestrup]

I did a quick look at man startx and didn's see this option, although I expected to. But even so, I wasn't sure I wasn't doing a major no-no by starting it manually.

Anyway, I don't think the SSH problem is with the firewall. I don't think webigail is correctly configured to accept ssh connections, since I can't log in either, but the firewall shows no problem.

[sps.livejournal.com]

My machine at work says that it takes server options after a --, cross-refs to XFree86(1), and that says that it takes args like :1. So unless they changed something, startx -- :1 should be a go. And no, I can't think of any reason not to start it manually unless perhaps you're already logged in and something gets grumpy about locks on your profile (an idiocy that has just recently shown up in both Gnome and 'Zilla - for Pete's sake, guys, what's even unusual about logging in twice???? Haven't we always done that under Un*x???? I hate the way the Windows dweebs are taking over Linux, I really do...! [No offense intended to any non-dweeb Windows users reading this.])

All this stuff is well into user-land, and X is supposed to work with all of its setup in a RAM-based database that is reloaded on login (though Gnome and KDM don't seem to understand this philosophy adequately). The display manager works like the old login mechanism: it gives you a prompt and launches your session, then fades into the background; and if you think of startx working like screen's front end - or Emacs considered as a shell, for that matter - (except that it grabs an X server instead of putting a terminal emulator on a pty) you won't go far wrong.

Really, all that's in the kernel is the back-end drawing mechanism, though I'll admit that the interface is pretty fat to support acceleration. And - at least logically - applications talk to the (gardware or software) X server through a socket, just like console applications talk to the the (hardware or software) terminal through a pipe.

The diagrams are all a billion times simpler than Windows, though I admit there are lots of cute li'l devils (and/or penguins) in the details....

...And if you don't have a graphical interface you should still be able to check out sshd in init.d and/or the xinet rc file, whatever it's called...

[swestrup]

Yes, well the documentation for startx on webigail explains the algorithm and the dozens of places it looks to find a graphic manager to start. No cross references.

If I do a 'startx -- -1' I get help messages to STDERR that I haven't figured out how to capture, so that I can read them before they scroll off of the screen.

Anyway, its all moot now.

[sps.livejournal.com]

That wasn't a typo. 'startx -- :1' - :1 as in display 1.

Anyway, if they 'fixed' the docs like that, it's plain stupid; the explanation of the parameters used to be in the first or second sentence.

[swestrup]

I didn't think it was. Way back when you said the option was either 'startx -- -1' or 'startx -- :1'

I tried both and both generated a list of acceptable options, but I couldn't capture it to read.

And now all startx says is that server options depend on which server the algorithm finds.

[sps.livejournal.com]

Did you try/know about Shift-PgUp (scroll the terminal window) and/or 2>&1 |less (which captures error output as well as normal output)?

[swestrup]

I didn't know about the first of those, and I can never remember the syntax for the second. I did some quick checking in the bash manpages, but they're huge, and I didn't see what I wanted right away. Like I said, it was a steambath in my office then and thinking was difficult.

(Its much nicer today!)

[sps.livejournal.com]

Yeah, the stuff about interactive terminal control is hard to find out. The latter is documented in the bash manpage, under 'redirection'.

I'm very upset at whoever invented the syntax... not least because it's inconsistent between 'foo 2>&1 | bar' and 'foo >bar 2>&1' ! (Pipes appear to be processed first even though they are written last. Good thing, ability-wise, bad thing, memory-wise.)