Self-interest August 1993

self-interest@lists.selflanguage.org

3 participants
4 discussions

Re: Problems with XOpenDisplay
by urs＠cs.stanford.edu 25 Aug '93

25 Aug '93

Here's what our current VM does (in xlibPrims.c): Display* XOpenDisplay_wrap(char *name, void *FH) { // XOpenDisplay fails if a signal is received during the call. // All signals except user interrupts are therefore blocked. SignalBlocker sb(true); Display* result = XOpenDisplay(name); if (result == 0) { prim_failure(FH, PRIMITIVEFAILEDERROR); return result; } XSetErrorHandler(xErrorHandler); XSetIOErrorHandler(xIOErrorHandler); // add this file descriptor to list of all open files FD_SET(ConnectionNumber(result), &activeFDs); return result; } The SignalBlocker blocks all signals except user interrupts (see abort.h) (yes, we're using SIGVTALARM and SIGIO as well). This definitely improves things, but sometimes it still fails. Don't ask me why. -Urs

1 0

Re: Problems with XOpenDisplay
by R James Noble 25 Aug '93

25 Aug '93

mcr(a)ccs.carleton.ca (Michael Richardson) writes: > I originally had problems with XOpenDisplay (and thus xlib display > open:, and thus viewManager open:) not working immediately. It would > usually work the second or third try. > James Nobel's navel code tried 25 times before giving up. Yes - I've had this problem for a while now (over a year, anyway), and I just loop around display: open for a while. I think I sent _something_ off to the bugs list when this first came out. Urs Hoelzle <urs(a)otis.stanford.edu> writes: > In case there are any X hackers out there who really know what's > going on, we'd certainly appreciate any suggestions for a real fix. Um, no, and ever since I found the workaround I stopped worrying about it :-) I guess the same applies at your end I've had a look in the latest Xlib (which seems to be a lot bigger than the last time I looked at it :-). The code seems to do five retries on ECONNREFUSED within the connect call during XOpenDisplay, but doesn't handle anything else in the connect, in particular EINTRs. kjx

1 0

Re: Problems with XOpenDisplay
by urs＠cs.stanford.edu 24 Aug '93

24 Aug '93

> [ XOpenDisplay fails, esp. when display is remote ] I think we never figured out the real source of the problem, but it seems that XOpenDisplay will fail if any interrupt occurs while it's waiting for a response from the server. We "solved" the problem by blocking signals around the call. [Self gets 50 or so timer interrupts / sec, so any connection taking longer than that has a hard time getting established...] Let me know if you need more details. In case there are any X hackers out there who really know what's going on, we'd certainly appreciate any suggestions for a real fix. -Urs

2 1

Problems with XOpenDisplay
by mcr＠ccs.carleton.ca 22 Aug '93

22 Aug '93

I originally had problems with XOpenDisplay (and thus xlib display open:, and thus viewManager open:) not working immediately. It would usually work the second or third try. James Nobel's navel code tried 25 times before giving up. I have since moved from being on the immediate subnet of the SparcStation I was running on, to being somewhat more remote... across campus. The good news is that I have my own NCD X terminal, and a portion of a little used Sparc II to play, the bad news is that I can no longer get XOpenDisplay to work. I'm wondering if the reasons behind the original problem had to do with some kind of non-blocking that was done to the socket? (for those curious about why I've moved, finger this account, or try: http://spiff.carleton.ca:8001/)

1 0

2024

2023

2022

2021

2020

2019

2018

2017

2016

2015

2014

2013

2012

2011

2010

2009

2008

2007

2006

2005

2004

2003

2002

2001

2000

1999

1998

1997

1996

1995

1994

1993

1992

1991

1990

Self-interest August 1993