[net.emacs] emacs crash - appeal for ideas

simard@loral.UUCP (Ray Simard) (12/04/84)

Hi there!

I have asked this before, but recevied no replies.  I'll try again.

Often, when a search-forward is executed and keys are accidentally
typed, emacs (Gosling flavor) goes inert for a long while, then
prints the message "Starting subshell" and does that.
Exiting or killing the shell locks up the terminal completely, requiring
killing emacs and/or the login shell from another terminal.  At that
time, often the checkpoint file contains such things as hunks of
the password file, or a list of my environment variables, and
other randomish garbage.  Sometimes the work is salvageable, and
sometimes not.

Any ideas as to why this happens, and how to avoid it (other than
the obvious "don't type keys while emacs is searching" ??

advTHANKSance

-- 

[     I am not a stranger, but a friend you haven't met yet     ]

Ray Simard
Loral Instrumentation, San Diego
{ucbvax, ittvax!dcdwest}!sdcsvax!sdcc6!loral!simard

...Though we may sometimes disagree,
   You are still a friend to me!

chris@umcp-cs.UUCP (Chris Torek) (12/07/84)

> Often, when a search-forward is executed and keys are accidentally
> typed, emacs (Gosling flavor) goes inert for a long while, then
> prints the message "Starting subshell" and does that.

> Any ideas as to why this happens, and how to avoid it (other than
> the obvious "don't type keys while emacs is searching" ??

In the keyboard reading code (probably in fill_chan or some variant
thereof depending on whose mods you have) you fill find something
like

	if (err)
	    return;

This is bogus; delete it.  Your problems should vanish.

(For anyone who cares, the problem is that there is a stdio-like
structure holding an input character count and a pointer.  If the
character count was positive before a --, the next character from the
pointed-to area is used, otherwise fill_chan is called.  If fill_chan
returns early (returning no particular value), interesting random
things happen until err somehow gets cleared.)
-- 
(This line accidently left nonblank.)

In-Real-Life: Chris Torek, Univ of MD Comp Sci Dept (301) 454-7690
UUCP:	{seismo,allegra,brl-bmd}!umcp-cs!chris
CSNet:	chris@umcp-cs		ARPA:	chris@maryland