From mboxrd@z Thu Jan  1 00:00:00 1970
Message-ID: <aee9645dc20ddefc190b82ad38844302@vitanuova.com>
To: 9fans@cse.psu.edu
Date: Wed, 25 Aug 2004 17:29:57 +0100
From: rog@vitanuova.com
In-Reply-To: <6667e8d6d26adf997f815e3bd4325d25@quintile.net>
MIME-Version: 1.0
Content-Type: text/plain; charset="US-ASCII"
Content-Transfer-Encoding: 7bit
Subject: [9fans] ftpfs bug
Topicbox-Message-UUID: d98fa75a-eacd-11e9-9e20-41e7f4b1d025

ftpfs has an annoying little bug.  actually it has at least two, one
of which is trivial, and i've just submitted a patch.  the other is
shown up by this program:

#include <u.h>
#include <libc.h>
void
main(void)
{
	int fd;
	fd = create("/n/ftp/blah", OWRITE, 0666);
	if(fd == -1)
		print("create: %r\n");
	else
	if(dirfstat(fd) == nil)
		print("fstat: %r\n");
}

not only does the fstat fail, but the failed fstat somehow causes the
newly created file to escape the cache - it only shows up when
reconnected to the ftp site.

the above isn't a problem usually, except that inside inferno, (and
exportfs too) a newly created file is fstat()ed to find out its new
qid.  this means that

	cp somefile /n/ftp

gets:

	cp: can't create /n/ftp/somefile: file does not exist

except the file does exist, and shows up next time (zero length, of
course).

is anyone out there familiar with ftpfs?  i've no time for fixing
anything right now, but this might turn out to be a really trivial
change...