Return-Path: doug@sonosphere.com
X-Sender: doug@light.lightlink.com
In-Reply-To: <199704180355.XAA27623@harvey.cyclic.com>
References: <01BC4B6F.6B38F840@pc.ice.com> (message from Tim Endres on
 Thu,	17 Apr 1997 20:39:16 -0400)
Mime-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Date: Fri, 18 Apr 1997 08:11:42 -0400
To: Jim Kingdon <kingdonc@harvey.cyclic.com>, info-cvs@prep.ai.mit.edu
From: Doug Wyatt <doug@sonosphere.com>
Subject: Re: CVS unix server and MAC client?

At 23:55 -0400 4/17/97, Jim Kingdon wrote:
> > 1)  CVS command line client with Apple's MPW
>
> Is there such a thing?  I haven't heard about it.  The CVS 1.9 Mac
> client is driven by AppleEvents (it also has a mode where it puts up a
> window, you type in a cvs command, and then the program exits, but
> without having tried it I would imagine that is probably useful mostly
> for debugging).

Matthias Neeracher, author of GUSI (a sockets interface for Mac), wrote an
MPW tool version of CVS in December.  It's in his FTP space ... a web
search for GUSI will probably lead you there.

It wasn't really clear to me how to use the MacCVS AppleEvent interface so
I ended up hacking a tiny stupid shell so I could have environment
variables.  And use the same command-line interface that the manual uses.

I also fixed a handful of bugs:

1) "release" with the '-d' option would fail.  This is because GUSI's
readdir() does not tolerate the file being deleted while the directory is
being walked through; files get skipped and not deleted.  I am a Unix
neophyte so I don't know whether this is a reasonable solution or a
horrible kludge, but a call to rewwinddir() at the end of the while() loop
in deep_remove_dir solved it.

2) In rcmd(), in rcmd.c, the value of errno after connect() was being being
destroyed by a call to close().  This would cause intermittent failures
when connecting (more often than not, in my case .... how did it ever work?
:) )

		sin.sin_len = sizeof(struct sockaddr_in);
		sin.sin_family = hp->h_addrtype;
		sin.sin_port = rport;
		bcopy(hp->h_addr_list[0], &sin.sin_addr, hp->h_length);
		if (connect(s, (struct sockaddr *)&sin, sizeof(sin)) >= 0)
			break;
		{
		int oerrno = errno;		// dsw added 4/97
		(void)close(s);
		errno = oerrno;			// dsw added 4/97
		}
		if (errno == EADDRINUSE) {
			lport--;
			continue;
		}
		if (errno == ECONNREFUSED && timo <= 16) {
		...

3) Another failure in "release" with the -d option was because
macos_fixpath was unprepared to handle "..", which the "release" code and
possibly other parts of the program pass to chdir().

char *
macos_fixpath (const char *path)
{
	char		*sepCh;

	strcpy( scratchPath, ":" );

	if( (*path == '.') && (*(path+1) == '/') )
		strcat( scratchPath, path+2 );
	else if (*path == '.' && path[1] == '.') {
		// dsw added handling of ".." 4/97
		strcat( scratchPath, "::");	// so now it's ":::"
									// stupid, but later the code skips the
									// first ':'
		if (path[2] == '/')
			strcat(scratchPath, path + 3);
	}
	else if( strcmp(path, ".") != 0 )
		strcat( scratchPath, path );
	while( (sepCh = strchr(scratchPath, '/')) != NULL )
		*sepCh = ':';

4) The console output will be *much* *much* faster with the addition of the
following lines in MacOS_Initialize():

	//SIOUXSettings.showstatusline = TRUE;  remove this line dsw 4/97
	setvbuf(stdout, NULL, _IOLBF, 512);	// line-mode buffering
	setvbuf(stderr, NULL, _IOLBF, 512);

For some reason Matthias couldn't remember, GUSI sets the console to
unbuffered output.

Oops, sorry about all the C++ - style comments.

Unfortunately, until my colleague John Cooper gets the NT CVS client
working right with our server, we're still using RCS.  It's weird, NT CVS
works fine here where the only real difference is that my TCP connection is
via Ethernet/ISDN instead of RAS.  But maybe we'll just use jCVS once the
Apple 1.1 Java VM comes out...

Doug


--
Doug Wyatt                      software development, Opcode Systems Inc.
work:     doug@opcode.com       http://www.opcode.com/
personal: doug@sonosphere.com   http://www.sonosphere.com/doug/
"Chaotic action is preferable to orderly inaction." - John Geirland