Subject: Re: A couple of suggestions for build.sh...
To: Luke Mewburn <lukem@netbsd.org>
From: Frederick Bruckman <fredb@immanent.net>
List: tech-toolchain
Date: 02/18/2003 17:38:25
On Wed, 19 Feb 2003, Luke Mewburn wrote:

> On Tue, Feb 18, 2003 at 09:04:16AM -0600, Frederick Bruckman wrote:
>   | I have a couple of suggestions for build.sh to make a build easier to
>   | restart after a failed attempt, for whatever reason.
>   |
>   | 1) Print the exact command line used to start the build before doing
>   | anything else, so it will be at the top of any captured log file:
>   |
>   | [...]
>
> I'm testing Andrew Brown's enhancement to this suggestion.

Cool.

>   | 2) I would like a "-c" or "-C" option to set NOCLEANDIR, and to not
>   | run any explicit "clean" targets. "-u" has "interesting"
>   | consequences when $DESTDIR is empty, and I'd like to not have to deal
>   | with that, when I only want to skip the "clean" phase (because I know
>   | the tree is already cleaned). It could be useful for kernel=, too.
>
> What are the `` "interesting" consequences '' of -u (UPDATE) on an
> empty $DESTDIR ?  (Fwiw: I use UPDATE all the time)

All the scripts (every file that's not generated, in fact), get
installed with the same modified time as the file in the checkout, or
at least, that's how it used to work. In the past, I have (foolishly)
removed all files older than a time-stamp file made before the build
started, and UPDATE isn't my friend anymore. I recall at least one
other user posting to the lists who had done exactly the same thing.
The failure mode when "true", "false", and "nroff" are missing is
puzzling, to say the least. Even if you never do anything that stupid,
it's nice to be able to scan through the directories with "ls -alt"
to see which files are obsolete.

Frederick