[2570] in SIPB_Linux_Development
/bin/login (ttylogin) and /etc/sendmail.cf (athena-rc)
daemon@ATHENA.MIT.EDU (Derek Atkins)
Wed Feb 10 19:56:15 1999
To: svalente@MIT.EDU
Cc: linux-dev@MIT.EDU
From: Derek Atkins <warlord@MIT.EDU>
Date: 10 Feb 1999 19:55:57 -0500
Sal,
I noticed that you're actually symlinking /etc/sendmail.cf and
/bin/login. As I noted previously, symlinking /bin/login is a BAD
IDEA (TM), especially if there is ever an update of _either_ package,
util-linux or ttylogin.
If there is an update of util-linux, it will write-through the symlink
and we will lose /bin/athena/login. This is certainly not what we
want, but it is certainly what will happen (I've tested this in the
past). Also, as the ttylogin scripts currently exist, if there is
ever an upgrade to ttylogin it will wind up restoring the original
/bin/login (%postun _IS_ run on upgrade!) -- you never test to make
sure /bin/athena/login still exists.
So, the current ttylogin package is broken in more ways than one. I
would _HIGHLY_ recommend you revert to the old ttylogin scripts, as it
tested for all these cases and would always DTRT except for _one_ case
(which is if you wind up running the trigger after a patch is created
but before you actually perform the upgrade)... It's a weird race
condition, but I actually _just_ figured that out this second.
I don't know whether sendmail.cf will do better. It probably is
installed as a config file, so making it a symlink should be ok
in terms of upgrades.
-derek
--
Derek Atkins, SB '93 MIT EE, SM '95 MIT Media Laboratory
Member, MIT Student Information Processing Board (SIPB)
URL: http://web.mit.edu/warlord/ PP-ASEL N1NWH
warlord@MIT.EDU PGP key available
