From mboxrd@z Thu Jan  1 00:00:00 1970
To: 9fans@cse.psu.edu
Subject: Re: [9fans] pptp.c install failed
From: "Russ Cox" <rsc@plan9.bell-labs.com>
MIME-Version: 1.0
Content-Type: text/plain; charset="US-ASCII"
Content-Transfer-Encoding: 7bit
Message-Id: <20011111213252.1487419A52@mail.cse.psu.edu>
Date: Sun, 11 Nov 2001 16:32:49 -0500
Topicbox-Message-UUID: 1e8fd914-eaca-11e9-9e20-41e7f4b1d025

> Why not ``rerrstr(char *buf, int size)''?
> Or even ``rerrstr(Str buf)''?
>
> C needs a counted array type, but until then we shouldn't
> cut corners.

As part of the 9P2000 changes, we fixed both errstr and
rerrstr as you suggest:

	int errstr(char *err, uint nerr)
	void rerrstr(char *err, uint nerr)

ERRMAX takes the place of ERRLEN, but it is a convention
rather than a specification: if you use one that is too big
or too small, the safety of your program isn't put in jeopardy.

Thus the code in question originally read:

	void
	ewrite(int fd, void *buf, int nbuf)
	{
		char e[ERRMAX], path[64];

		if(write(fd, buf, nbuf) != nbuf){
			rerrstr(e, sizeof e);
			strcpy(path, "unknown");
			fd2path(fd, path, sizeof path);
			myfatal("write %d to %s: %s", nbuf, path, e);
		}
	}

which I think you wouldn't have any problems with.

I did think about posting an rerrstr(char*, uint), but
our rerrstr uses utfecpy to make sure it truncates at
a character boundary, and that function is post-distribution
too, so I stopped the cascade by assuming ERRLEN: since
the distribution errstr already does it, I don't see a problem
doing it in other places until we get a proper distribution out.

Russ