home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: vmsnet.uucp
- Path: sparky!uunet!munnari.oz.au!comp.vuw.ac.nz!rata.vuw.ac.nz!don
- From: don@rata.vuw.ac.nz (Don Stokes)
- Subject: Re: system abort error
- Nntp-Posting-Host: rata.vuw.ac.nz
- Message-ID: <C09I0z.I5F@comp.vuw.ac.nz>
- Organization: Victoria University of Wellington CSC
- Sender: news@comp.vuw.ac.nz (News Admin)
- References: <1993Jan2.160213.1057@cmkrnl.com>
- Date: Sun, 3 Jan 1993 05:02:58 GMT
- Lines: 39
-
- jeh@cmkrnl.com writes:
- > don@rata.vuw.ac.nz (Don Stokes) writes:
- > > .... which is the result of outstanding $QIOs being $CANCELled when
- > > the protocol is torn down; there were some circumstances before this was
- > > done where UUCICO could hang if carrier was unexpectedly lost. The
- > > SS$_ABORT tels you it's not suffering that bug. 8-)
- >
- > Exactly!
-
- Yeah, I was "supporting" a site that had a particularly flakey connection
- and could reproduce this problem pretty much at will (ie by waiting a day
- or two for it to happen -- I'd seen it happen elsewhere once in a blue moon).
-
- What I _think_ was happening was that the modem was dropping CTS when it
- hung up, and the terminal driver was duly holding up traffic; uucico would
- correctly detect the loss of carrier, but didn't take any action (it still
- doesn't -- just reports the fact), continuing to Q IOs. It would eventually
- time out, but for some reason would hang, presumably when it tried to do a
- syncronous I/O (SETMODE!HANGUP QIO?) behind the wedged ones. (I
- haven't looked at the code for a while.) Note that an incoming call to the
- modem would raise the necessary signals again, and uucico would go merrily
- on its way....
-
- Anyway, blowing all outstanding I/Os out of the water before attempting to
- close the connection down has fixed it. Kudos to Mark (or was it Tom --
- bit rot is setting in) for putting max effort into fixing it -- those sorts
- of bugs are pigs to track down.
-
- > (One wonders just how many centuries of VAX cpu time has been spent checking
- > IOSBs over the past 15 years. Not to mention probing buffers, copying
- > buffers, etc.)
-
- Not as many as have been burned throughout the C-speaking world looking for
- the ends of nul-terminated strings.... <sigh> K&R, you botched. 8-(
-
- --
- Don Stokes, ZL2TNM (DS555) don@zl2tnm.gen.nz (home)
- Network Manager, Computing Services Centre don@vuw.ac.nz (work)
- Victoria University of Wellington, New Zealand. +64-4-495-5052
-