Subject: Re: "make install" lossage, redux
To: None <firstname.lastname@example.org>
From: Greg A. Woods <email@example.com>
Date: 12/28/1999 13:47:03
[ On Saturday, December 25, 1999 at 18:22:27 (-0700), Miles Nordin wrote: ]
> Subject: Re: "make install" lossage, redux
> this makes sendmail into a ``host'' binary, so in a cross build you'd have
> to build a sendmail for the host architecture just to run this command,
> and then the usual sendmail for the target as well.
> but if the db files have endiness, it's completely hopeless anyway, yes?
However, in the case of sendmail in particular, and perhaps in the case
of all other .db files too, it is completely and totally bogus to be
building them in the "make install" phase. These things should only
ever be created on the final target host (i.e. either by sysinstall, or
by some manual process.
I suppose if $DESTDIR == / and if you really do a "make distribution"
then it would be OK to expect sendmail to be installed already and to
use it to build /etc/aliases.db from the just installed /etc/aliases.
In fact in the specific case of sendmail's /etc/aliases.db file, is
there ever any need to create the .db file any more -- doesn't sendmail
now have an option to always automatically rebuild the aliases file on
Ideally all programs should function adequately if their .db files are
missing or out of date... (albiet possibly slowly and perhaps not in a
totally identical way if the data is somehow differently percieved than
it would be in a db file, such as with a duplicate key).
Greg A. Woods
+1 416 218-0098 VE3TCP <firstname.lastname@example.org> <robohack!woods>
Planix, Inc. <email@example.com>; Secrets of the Weird <firstname.lastname@example.org>