home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!noc.near.net!chaos!random.ccs.northeastern.edu!agathos
- From: agathos@ccs.northeastern.edu (Michael W Wojcik)
- Newsgroups: comp.unix.aix
- Subject: Re: XGKS on AIX RS6000 problem
- Message-ID: <1992Jul23.144512.12985@random.ccs.northeastern.edu>
- Date: 23 Jul 92 14:45:12 GMT
- References: <Jul20.055403.9171@yuma.ACNS.ColoState.EDU>
- Sender: news@random.ccs.northeastern.edu
- Organization: Comp. Sci. @ NU
- Lines: 20
-
- I worked on IBM's original XGKS (first under AOS [BSD 4.3] on the RT, if
- anyone's curious). I don't remember the particular problem you described,
- but I do recall that input -- mouse and keyboard -- gave us a lot of
- trouble, both in the original implementation and when we first tried it
- on the 6000. XGKS (on the RT) sets the X display server socket to signal
- (FIOASYNC ioctl), traps SIGIO, and peeks the message queue; that caused
- some problems (eg. some Xlib calls would bomb if an event arrived while
- they were executing due to reentrancy issues, so we had to sigpause
- around them). Maybe it's interfering with xterm's message handling somehow.
-
- One possibility might be to fork, have the child do the XGKS (close its
- stdin before starting XGKS, maybe), and have the parent do the stdio stuff.
- The two could communicate via a pipe or something. It's a bit extravagant,
- but it might work, and it doesn't require hacking the XGKS code.
-
- Maybe one of the other original XGKS people can give you more help (Dave?
- Bruce? Amy? where is everyone?). Isn't IBM still supporting the thing?
-
-
- Michael Wojcik
-