From mboxrd@z Thu Jan 1 00:00:00 1970 From: Josh Marshall Subject: Re: [EXTERNAL]Re: Modifying squashfs usage for `guix pack` for supporting singularity Date: Fri, 13 Mar 2020 16:48:20 +0000 Message-ID: References: <87v9n9smtp.fsf@gnu.org> <87ftedr0hz.fsf@inria.fr> , <87r1xwnrac.fsf@inria.fr> Mime-Version: 1.0 Content-Type: text/plain; charset="windows-1252" Content-Transfer-Encoding: quoted-printable Return-path: Received: from eggs.gnu.org ([2001:470:142:3::10]:47365) by lists.gnu.org with esmtp (Exim 4.90_1) (envelope-from ) id 1jCnUA-0001Ek-NF for guix-devel@gnu.org; Fri, 13 Mar 2020 12:48:37 -0400 Received: from Debian-exim by eggs.gnu.org with spam-scanned (Exim 4.71) (envelope-from ) id 1jCnU8-00064Z-1O for guix-devel@gnu.org; Fri, 13 Mar 2020 12:48:33 -0400 Received: from esa2.hc3102-66.iphmx.com ([139.138.57.83]:60691) by eggs.gnu.org with esmtps (TLS1.0:DHE_RSA_AES_256_CBC_SHA1:32) (Exim 4.71) (envelope-from ) id 1jCnU7-0005vG-Me for guix-devel@gnu.org; Fri, 13 Mar 2020 12:48:31 -0400 In-Reply-To: <87r1xwnrac.fsf@inria.fr> Content-Language: en-US List-Id: "Development of GNU Guix and the GNU System distribution." List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Errors-To: guix-devel-bounces+gcggd-guix-devel=m.gmane-mx.org@gnu.org Sender: "Guix-devel" To: =?Windows-1252?Q?Ludovic_Court=E8s?= Cc: "guix-devel@gnu.org" Glad to have contributed :) ________________________________________ From: Ludovic Court=E8s Sent: Friday, March 13, 2020 12:38 PM To: Josh Marshall Cc: guix-devel@gnu.org Subject: Re: [EXTERNAL]Re: Modifying squashfs usage for `guix pack` for sup= porting singularity Hi Josh, Josh Marshall skribis: > While the images seem to work well in most cases, removing ways for bits = to change seems to be in the spirit of guix, and so the "-all-time" and "-m= kfs-time" options remove one source of change. Additionally, without the "= --no-recovery" an extra file is made which is not useful to the process, an= d for my manual testing seemed to cause write or permission issues which co= uld cause failures. Guix already handles the failure cases in mksquashfs w= hich the recovery file is for and so the redundant work is not helpful, can= cause problems, slows down the process, and requires more resources. Oh, I hadn=92t understood that there were reproducibility issues with =91guix pack -f squashfs=92, sorry about that! Commits 24fb0dc0ab34ebb49509a3d5b4d84d8488670807 and b829864d747b3b24ef37cafe36e889527b060d4d implement what you suggest. I can confirm that now something like: guix pack -f squashfs sed --rounds=3D2 passes, which was not the case before. Thanks! > As for the information given by `guix pack --list-formats`, I'd change to= something like the following: > > ``` > tarball Self-contained tarball, ready to run directly on any Linu= x machine > squashfs Container suitable for Singularity 2.x and 3.x > OCI OCI compliant tarball ready for 'docker load' and usable by S= ingularity 3.x > sif Unsupported; container preferred by Singularity 3.x We cannot change the format name, for compatibility reasons, but I=92ll double-check that what =91-f docker=92 produces is indeed OCI, and if so, add it to the description. > For more information, refer to https://guix.gnu.org/manual/en/html_node/I= nvoking-guix-pack.html I don=92t think we=92re going to add references to the manual in --help and similar: it=92s always implicit that documentation is in the manual. However, perhaps we could add terminal hyperlinks to make it more discoverable. Thanks, Ludo=92. --- The information in this email, including attachments, may be confidential a= nd is intended solely for the addressee(s). If you believe you received thi= s email by mistake, please notify the sender by return email as soon as pos= sible.