--On Wednesday, March 10, 2004 10:43 PM +0100 "Jostein Chr. Andersen"
<jostein@josander.net> wrote:
> 1. If this patch, for any reason, should be approved (well it is
> committed, so it is), then it should be merged to the 1.0.x branch as
> well. I'll be happy to vote for it in the STATUS file(even if I like
> suggestion 2 below better).
Well, we have *had* a discussion on the list about this - see 'Fixing the
GNU tar requirement' and 'Produce POSIX.1 compliant tarballs.' This patch
should be approved and why I did so and asked Ben to commit to trunk is
because we should not expect our end users to have GNU tar installed.
(Solaris's tar is the predominate one that we'll fail on right now.)
As for why it hasn't been proposed yet into 1.0.x was because I was
concerned that switching over to cpio this close to release could cause a
problem. However, in a followup on that thread, we've been soliciting
confirmation that the cpio-generated tarballs are still valid everywhere.
So far that seems to be the case, but I would not want to switch *for
1.0.1* without first ensuring that we haven't locked any platforms out by
doing this.
I would not be pained if 1.0.1 still had this fault and we resolved it in
1.0.2 if we don't receive enough confirmation that cpio won't mess us up.
However, the RM has final authority on how they produce the build - if they
want to use cpio for 1.0.1, cool; if not, no sweat. 1.0.2, at the least,
should provide a resolution.
> 2. Revert it. KISS! ;-)
Given that we're *actually* hitting this 100-character limit which produces
corrupted dirs on Solaris, I don't think that's a viable option. -- justin
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Wed Mar 10 23:01:49 2004