Path: utzoo!utgpu!news-server.csri.toronto.edu!rpi!usc!cs.utexas.edu!bcm!lib!thesis1.med.uth.tmc.edu
From: dfenyes@thesis1.med.uth.tmc.edu (David Fenyes)
Newsgroups: comp.unix.shell
Subject: Re: ash internals: evalstring() calls
Message-ID: <4892@lib.tmc.edu>
Date: 30 Mar 91 22:00:02 GMT
References: <4879@lib.tmc.edu> <1991Mar26.212622.757@am.sublink.org>
Sender: usenet@lib.tmc.edu
Organization: University of Texas Medical School at Houston
Lines: 40
Nntp-Posting-Host: thesis1.hsch.utexas.edu

In article <1991Mar26.212622.757@am.sublink.org> alex@am.sublink.org (Alex Martelli) writes:

>Here's the relevant portion from mash, my own ash dialect:
>
>		handler = &jmploc;
>		TRACE(("putprompt(%s)\n", prompter));
>		if (setjmp(jmploc.loc)) {
>			TRACE(("putprompt error recovery\n"));
>			popstackmark(&smark);
>			popfile();
>			p = 0;
>		} else {
>			setstackmark(&smark);
>			evalstring(prompter);
>			popstackmark(&smark);
>			p = lookupvar("__P");
>		}
>
>where "prompter" is a string I build up by catting __P=" in front of
>lookupvar("PS1"), and " after that.

My solution was very similar looking-- I call a function cheapexpand()
from putprompt in parser.c, which epands the ~/, then calls expandstr,
which cats up a similar string & calls evalstring in eval.c.

The problem I had was, of course, that syntax errors invoked error(),
which did a longjmp back to the main loop, which tried to parse the
offending string again.  Judicious local use of setjmp, INTON, and INTOFF
fixed the problem.

For those of us who use ash, it may be useful to pool information,
such as bug fixes, etc.  2 bugs I found are a bomb on an empty subshell
(), avoided by making an empty subshell a syntax error (I know it's a
kludge), and improper shell exiting from scripts in the -I mode, fixed
by including a `!top ||' in the test for exit on an EOF.

Bye,
--
David Fenyes                                 dfenyes@thesis1.hsch.utexas.edu
University of Texas Medical School           Houston, Texas