Subject: Re: New checkflist mechanism too strict ?
To: Luke Mewburn <lukem@wasabisystems.com>
From: Patrick Welche <prlw1@newn.cam.ac.uk>
List: tech-toolchain
Date: 07/22/2002 14:38:25
On Mon, Jul 22, 2002 at 10:11:32PM +1000, Luke Mewburn wrote:
> On Mon, Jul 22, 2002 at 09:53:21AM +0100, David Laight wrote:
>   | On Mon, Jul 22, 2002 at 10:17:41AM +0200, Xavier HUMBERT wrote:
>   | > Even if I don't update my userland every day, I regulary build into
>   | > /usr/build. And I build a distriobution (-d flag), in order to have
>   | > up-to-date /etc scripts (security, rc.blah, etc.)
>   | > 
>   | > Since "checkflist" mechanism has been introduced, make distribution fails
>   | > every 2 or 3 other builds, on some manpage, I presume.
>   | 
>   | Yes...  I noticed that,
>   | Fortunately it is the last thing the build (currently) does.
>   | I guess you could add a '-' to the front of the command in the makefile.
>   | Maybe that check ought to be optional?
> 
> No, it should not be optional.
> 
> The addition of "make checkflist" was added to the "maketars" target in
> distrib/sets/Makefile to ensure that the generated tar files have all
> the appropriate contents.  In Xavier's case, his /usr/build appears to
> have a couple of now-obsolete files.
> 
> The "sets" target in distrib/sets/Makefile depends upon "maketars",
> and is only called in the top-level Makefile or etc/Makefile by
> "make release" or "make snapshot".
> 
> AFAICT, a "./build.sh -d DESTDIR" invokes "make distribution", so it
> shouldn't even invoke checkflist-as-part-of-maketars.  It's only if
> you add "-R RELEASEDIR" to the build.sh invocation that it should.

Not that I've checked, but what of a release built with eg MKKERBEROS=no
so eg kinit just wouldn't be there? Is it just something to ignore as then
it isn't a standard release?

Cheers,

Patrick
(PS thanks for adding the bsd.own.mk to acorn32/stand/Makefile :-) )