rbj@DSYS.ICST.NBS.GOV (Root Boy Jim) (04/05/89)
Just the other day I wrote: ? I'm running X11R3 + patches 1-9, Sun 3/180, SunOS 3.5 + nameserver kit. ? Awhile back I posted a message about problems with xlsfonts dying... Perhaps I didn't make myself clear, I need a response. The last time I asked about this, I got many replies. Have I used up my quota of stupid questions, or are all the gurus on vacation? ? I remember seeing something about `if the client can't keep up with the ? server, the server may get impatient, decide that the client doesn't ? exist anymore, and clean up its end, thereby breaking the connection'. ? I found that somewhere around 4K (pipe size limit?) was about all I ? could get out of xlsfonts without it dying or being killed. ? Are these problems related? And, yet another manifestation: when starting up xinit alone with no initialization files, cat'ing an X Makefile kills the xterm (and naturally the server) with the message: xterm: Invalid argument I have grepped the entire sources (*.[chlys]) and can't find who prints out that message and why. Must be somewhere tho. Is there a way to trace/log all packets to and from the server? ? Catman Rshd <rbj@nav.icst.nbs.gov> ? Author of "The Daemonic Versions"
rws@EXPO.LCS.MIT.EDU (Bob Scheifler) (04/05/89)
Perhaps I didn't make myself clear, I need a response. I never thought otherwise. Have I used up my quota of stupid questions, or are all the gurus on vacation? The gurus are all exhausted. ? I found that somewhere around 4K (pipe size limit?) was about all I ? could get out of xlsfonts without it dying or being killed. We don't see this problem on various OSes here (we don't have SunOS 3.5). We have various bug reports for various OSes saying the server is failing to handle their OS's particular braindamaged error return for the write(v) system call when various preconceived kernel implementor's notions of how the call should work aren't met. It's hard to believe there are so many different semantics applied to this call, and so many weird interpretations placed on errors (e.g., returning EWOULDBLOCK is the data is "too big"; geez). (Now add SYSV signals and stir ...) It's almost enough to make one want to attend a POSIX meeting. Is there a way to trace/log all packets to and from the server? Use xscope.
rbj@DSYS.ICST.NBS.GOV (Root Boy Jim) (04/07/89)
? From: rws@expo.lcs.mit.edu (Bob Scheifler) ? The gurus are all exhausted. Understandable :-) ? ? I found that somewhere around 4K (pipe size limit?) was about all I ? ? could get out of xlsfonts without it dying or being killed. ? We don't see this problem on various OSes here (we don't have SunOS 3.5). We ? have various bug reports for various OSes saying the server is failing to ? handle their OS's particular braindamaged error return for the write(v) system ? call when various preconceived kernel implementor's notions of how the call ? should work aren't met. It's hard to believe there are so many different ? semantics applied to this call, and so many weird interpretations placed on ? errors (e.g., returning EWOULDBLOCK is the data is "too big"; geez). (Now ? add SYSV signals and stir ...) It's almost enough to make one want to attend a ? POSIX meeting. Or perhaps an X/Open meeting, or if you're real desparate, OSF :-) I have more information on what my problem is; UNIX domain sockets. If I set my DISPLAY to dsys:0 it works fine, but unix:0 dies. If I use localhost:0 it works too. Thanks for all the help. Catman Rshd <rbj@nav.icst.nbs.gov> Author of "The Daemonic Versions"
pinkas@hobbit.intel.com (Israel Pinkas ~) (04/10/89)
In article <8904061829.AA07362@dsys.icst.nbs.gov> rbj@DSYS.ICST.NBS.GOV (Root Boy Jim) writes: > I have more information on what my problem is; UNIX domain sockets. > If I set my DISPLAY to dsys:0 it works fine, but unix:0 dies. > If I use localhost:0 it works too. I saw something like this under X10 when running YP. This happened on Suns with xterm. There was a patch for xterm that fixed this. I'll see if I can find it and send it to you. The problem, as I remember it, was that YP left an extra open file decriptor around. For some reason, xterm assuped that the pip it used to communicate with the process had a fixed file descriptor. You might also be able to find the patch in one of the earlier patches released for X11. Check for anything that patches xterm. The reason that localhost:0 works is that localhost is a synonym for your machine, using the TCP loopback. It is better that using `hostname`:0, but not as good as unix:0. (Many implementation will not put localhost traffic on the ethernet wire, but will put `hostname` packets out there for everybody to see.) -Israel Pinkas -- -------------------------------------- Disclaimer: The above are my personal opinions, and in no way represent the opinions of Intel Corporation. In no way should the above be taken to be a statement of Intel. UUCP: {amdcad,decwrl,hplabs,oliveb,pur-ee,qantel}!intelca!mipos3!cadev4!pinkas ARPA: pinkas%cadev4.intel.com@relay.cs.net CSNET: pinkas@cadev4.intel.com