home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: comp.os.linux
- Path: sparky!uunet!usc!sol.ctr.columbia.edu!ira.uka.de!uni-heidelberg!fht-mannheim!aix01.rz.fht-mannheim.de!mrex
- From: mrex@aix01.rz.fht-mannheim.de (Martin Rex)
- Subject: Xlinux.8514: 0.97-1 ok -- 0.97-4 not (long)
- Message-ID: <5srna_#_@aix01.rz.fht-mannheim.de>
- Date: Tue, 08 Sep 92 16:14:53 GMT
- Organization: Fachhochschule fuer Technik Mannheim
- X-Newsreader: Tin 1.1 PL4
- Lines: 93
-
- Hi,
-
- I have recently installed Linux 0.97pl1 on my machine
- (i486/33/20MB/200MB AT-Bus/ATI Graphics Ultra), and GCC-2.2.2d,
- X-1.1 and Kevin Martin's Xlinux.8514.
-
- Under 0.97pl1 everything was working fine, and I have successfully
- compiled about 20 X-programs so far. The only bad thing about
- compiling: I HAVE TO use gcc-2.2.2 for the linking, when compiling
- X-Software (linking with X-libs), otherwise I get SegVs. :-((
- (using shared libraries; statically linked seems to work in
- rare occasions.)
-
- The last weekend I have experienced file system corruption twice
- on my main minix partition (50 MB), and spent about 10 hours of
- saving and restoring (and verifying as good as possible) the data
- on this partition. Yes, I have used "strip" from the 0.97pl1
- rootdisk, so I guess that was one reason from what I understand
- from Linus' posting about patch4.
-
- I've been heavily using X on 0.97pl1, and it works really well.
-
- Yesterday I have ftp'ed the 0.97pl4-kernel and comiled it on
- my machine (it takes less than 7 minutes from scratch), and
- tried X on it. It starts up and it runs, but I have big
- problems on quitting X! The CTRL-ALT-BACKSPACE works, but
- when I end my startup-shell (I have "exec xterm -ls" as the
- last line of my xinitrc), the X-Server doesn't close down
- correctly and hangs.
-
- When I kill all programs that use the X-server ("xclock, twm"),
- and then kill the xterm as the last program accessing the server,
- then the Server shuts down smooth and returns to the virtual console.
- As soon as there is another program using the X-Server when the
- xinitrc-process (my startup shell/xterm) is killed/ended,
- the keyboard and screen freezes with a partially cleard display.
- (the system is still operational through the serial port then.)
- When this once happened, I am not able to restore normal behavior
- on the "console". I may kill every X-related process from the serial
- session, but the screen stays the same. When I restart X from the
- serial session, the Display is back under control of the X-server,
- and even the mouse is working, but the keyboard is not recognized
- anymore. When I kill this renewed session first twm & xclock, then
- the startup xterm/shell, the screen returns to text-mode,
- but the keyboard still doesn't work.
-
- I have never had any such problem with 0.97pl1, where I could exit
- the startup-shell with as many as 8 windows open, and the X-Server
- would still shut down smoothly.
-
- But there's something else I've noticed under both pl1 & pl4 which
- has been mentioned in this list recently. Sometimes after X-shutdown
- a zombie xinitrc stays around, and accumulates huge amount of
- process time. It usally drains about 30-50% of the power of the
- machine. I usually notice it only when comiling or having xfishtank
- in the background, normally the machine is still acceptably responsive.
- The zombie xinitrc always has a <defunct> brother, so I think
- there is some kind of deadlock of two processes.
-
- During the little time of trying last night, I also had a very strange
- behaviour of 0.97pl4. After killing and rerunning X serveral times
- over the serial session trying to get back a usable console,
- the machine must have trashed/mangled or filled up its memory and
- page space, because it was suddenly very slow with its responses,
- and it was accessing the hard disk every several seconds, but there
- weren't any process running which could cause this harddisk activity,
- I had only started X (startx), but this time it could not even get
- the control back over the screen. The machine was really really slow,
- and the strange disk activities went on for over 10 minutes
- (until I hit the hard-reset). I was trying to get the control back,
- but my serial session fell asleep on an "ps" command (it listed
- only the first three processes very very slowly). I pulled the
- serial connection because there was no further reaction to Crtl-Cs
- and Ctrl-Zs. After reconnection of the serial cable, it took about
- one minute before the login prompt reappered, and then the characters
- were slowly echoed back. I even got the login greeting, but I didn't
- get to the shell prompt (maybe my login script was to long -- the machine
- still was down to the knees with workload -- but I have no idea what
- it was doing! I assume it was mainly swapping (the disk activity ...)
- (btw. my swap-partition is 10MB).
-
- I'd rather stay with 0.97pl1, if there's something wrong with pl4 (pl3),
- but I would like to get the file system fixes in my pl1 at least!
-
- Any ideas ?
-
- -Martin
-
- --
- Martin Rex | "Have the courage to take
- Rechenzentrum | your own thoughts seriously,
- Fachhochschule fuer Technik Mannheim | for they will shape you."
- E-Mail: mrex@aix01.rz.fht-mannheim.de | -- Albert Einstein
-