[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <JhK61rXiXRkJbVJqHH9kRlLM_zO-J6fPM-NCa2P1eKSIfXzpunRtwJNMS4fliDWqMBhQKqp5t3fmUmKLhuSAeqTS6nVogdqnVyxagsH2z9M=@protonmail.ch>
Date: Wed, 09 Oct 2019 13:56:52 +0000
From: Dmitry Goldin <dgoldin@...tonmail.ch>
To: Dmitry Goldin <dgoldin@...tonmail.ch>
Cc: Masahiro Yamada <yamada.masahiro@...ionext.com>,
Andreas Schwab <schwab@...ux-m68k.org>,
Greg KH <gregkh@...uxfoundation.org>,
"linux-kernel\\\\\\\\\\\\\\\\@vger.kernel.org"
<linux-kernel@...r.kernel.org>,
"joel\\\\\\\\\\\\\\\\@joelfernandes.org" <joel@...lfernandes.org>,
Ben Hutchings <ben@...adent.org.uk>
Subject: Re: [PATCH] kheaders: substituting --sort in archive creation
Hi again,
Sorry for the delay, I'm currently traveling and only have access to a
rather weak machine so compiles took ages.
A few remarks regarding this patch;
This version is not fully identical to the previous invocation, as the
sort order differs a little, but I tried to get it as close to the
original as I was able to. Unfortunately no sort flag seems to quite
replicate tars sorting.
Some noteworthy details in this version are `./` in the find format string, LC_ALL
for sort and --no-recursion;
The format string is in place to replicate the exact behaviour of the previous
invocation, which included a leading `./`. I don't mind dropping this,
if anyone feels strongly about it, but initially I set out to reproduce
the result as closely as possible.
LC_ALL=C is required as locale can impact sort order.
--no-recursion is required to prevent duplication in the resulting archive,
because both, folders and files, are already supplied from find and sort.
I checked this part of the script on an old debian lenny (5.0.10) system and it
behaves as expected, except for xz support not being available in
the shipped gnu tar (v1.20).
As far as other testing goes, I have compiled 5.3.5 on NixOS with kheaders as a
module where both runs produced identical results.
Because of the limited computational power available to me right now, I
did not have time yet to try the baked-in version or any additional
compile runs. I would appreciate if reviewers could do a few runs too.
Andreas: Could you give this patch a try and see if this works for you?
--
Thanks and best regards,
Dmitry
Powered by blists - more mailing lists