On Mon, 4 Oct 1999, Jeremy C. Reed wrote:
> Here is my debugging output (JUST the "Local deliveries" section)
> (the folder /var/spool/virtual/poptest.reedmedia.net does NOT exist,
> but when it does exist, it works)
That debugging output looks exactly as if you haven't got
create_directory set at all. Mysterious. Can you check the setting with
exim -bP transport virtual_localdelivery
to ensure it really is there? It looks as if code hacking is needed.
Around line 1083 in the file src/transports/appendfile.c there are the
lines
/* If ob->create_directory is set, attempt to create the directories in
which this mailbox lives. We know we are dealing with an absolute path,
because this has been checked above. */
if (ob->create_directory)
{
char *p = strrchr(path, '/');
*p = '\0';
if (directory_make("/", path, ob->dirmode, FALSE) < 0)
{
addr->basic_errno = errno;
addr->message =
string_sprintf("failed to create directories for %s: %s", path,
strerror(errno));
DEBUG(9) debug_printf("%s transport: %s\n", tblock->name, path);
return;
}
*p = '/';
}
Before those lines you could insert
DEBUG(9) debug_printf("ob->create_directory=%d\n", ob->create_directory);
and check that it has the value 1. It might also be worth putting in
something just before the *p = '/'; line just to check that it is
getting there.
The actual directory building is done in a function whose code is in
directory.c ... ... Aarrgghh! I've just spotted a bug in the code while
checking that. It returns FALSE if making the directory fails, but the
code in appendfile is checking for a negative return.
OK. The problem is that making the directory is failing for some reason,
but that isn't being reported correctly. Please change the line in
appendfile (part of the block shown above) that reads
if (directory_make("/", path, ob->dirmode, FALSE) < 0)
to
if (!directory_make("/", path, ob->dirmode, FALSE))
It should then report the error it is getting, and we should be a bit
further along the road to figuring out what is going on.
Sorry about that. That is *very* old code, so you have just helped to
kill a bug of great longevity.
Regards,
Philip
--
Philip Hazel University of Cambridge Computing Service,
ph10@??? Cambridge, England. Phone: +44 1223 334714.
