Subject: Re: Dynamic libraries, bad libc and problems...
To: Johnny Billquist <email@example.com>
From: Bill Studenmund <firstname.lastname@example.org>
Date: 10/01/2002 16:41:40
On Tue, 1 Oct 2002, Johnny Billquist wrote:
> On Tue, 1 Oct 2002, Bill Studenmund wrote:
> > ??? Why do that? A much safer thing to do is make a release (add -R and
> > don't have -d around) and install the tar files.
> *That* I would probably recommend. Then you get files installed in what
> would probably be defined as the very wrong order. In case your tar stops
> somewhere in the middle, you'd have a really botched system, such as /bin
> installed, but not /libexec, or /lib.
> You should be able to guess what happens after that.
Uhm, why is your tar going to stop in the middle? tar tzf <file> will
check and see if the file is ok or not, if you're that concerned.
If you don't like tar, you can use pax.
Also, think about how many processes get started during untarring vs how
many get started in the build-to-root method. For tar, it is zero (yes,
tar gets forked, and gunzip gets forked, but that's before the untarring
itself starts). For the build method, it's a lot. Each install is a fork,
and there are a number of make invocations along the way. Also there could
be a number of other programs invoked along the way too.
I really fail to see how an install method that does all of the library
updates in a manner that lets us quiesce process forking/execing is less
desirable to one that relies on a fair number of fork/execs.