home *** CD-ROM | disk | FTP | other *** search
- Xref: sparky comp.sys.next.misc:23526 comp.sys.next.programmer:7994
- Path: sparky!uunet!ogicse!das-news.harvard.edu!spdcc!merk!roscom!monty
- From: monty%roscom@think.com (Monty Solomon)
- Newsgroups: comp.sys.next.misc,comp.sys.next.programmer
- Subject: Re: Bug in Kermit5A(188) under 3.0 (with a patch)
- Message-ID: <1993Jan5.200524.11251@proponent.com>
- Date: 5 Jan 93 20:05:24 GMT
- Article-I.D.: proponen.1993Jan5.200524.11251
- References: <ntomczak.725563464@vega>
- Sender: monty@proponent.com (Monty Solomon)
- Reply-To: Monty Solomon <roscom!monty@think.com>
- Followup-To: comp.sys.next.misc
- Organization: Proponent
- Lines: 48
-
- In article <ntomczak.725563464@vega> ntomczak@vega.math.ualberta.ca (N
- Tomczak-Jaegermann) writes:
- >
- > There seems to be a bug in the latest version of Kermit, 5A(188),
- > 23 Nov 92, from columbia.edu. At least as far as NeXT is concerned.
- > When you will try to open a new file, either by specifying a name for a
- > received file different from an automatic default or by using an
- > explicit 'open' Kermit command then most likely you will get
- > "Access denied" or an equivalent message.
- >
- > The problem is that Kermit should be owned by uucp to get an access to
- > communication ports and lock files and runs as "suid program" so you
- > can use it from a regular user account (it will refuse to run if owned
- > by a root). For some reasons Kermit swaps internally real and
- > effective id's ...
-
-
- Here is a reply from Frank da Cruz.
-
- Date: Tue, 5 Jan 93 9:31:47 EST
- From: Frank da Cruz <fdc@watsun.cc.columbia.edu>
- Subject: Re: FYI - Bug in Kermit5A(188) under 3.0 (with a patch)
-
-
- The reason is that when you have a set-UID program, you want it to run with
- the user's real ID when it is accessing ordinary files and creating forks, not
- the set-UID. The set-UID-ness in Kermit is only for creating the lockfile and
- accessing the communication device (an artifact of a hideous design flaw of
- UNIX that has haunted us forever, namely that access to serial devices like
- communication ports is shared by default, rather than exclusive). See the
- installation instructions, ckuker.doc, for a fuller explanation.
-
- > ... and this is causing 'access()' library function from
- > NeXTSTEP 3.0 to fail when probing your directories. The following
- > patch seems to make things OK (at least for me and so far :-)):
- >
- Your patch is essentially correct. I regard it as a workaround for a bug in
- NeXT's implementation of access(). We discovered the same thing here last
- week and issued approximately the same patch. Updated sources and makefile
- have been installed in kermit/b on watsun.cc.columbia.edu, and a bug report
- was sent to NeXT.
-
- - Frank
-
-
- --
- # Monty Solomon / PO Box 2486 / Framingham, MA 01701-0405
- # monty%roscom@think.com
-