TL;DR: Why does jigdo break on old (3.1_r8) Debian ISOs, on tar.gz files that changed? Considering those files change, but the contents are identical or almost identical.
Hi, I was playing around with an old Debian version, and trying to get a copy of the DVD source ISO of Debian 3.1_r8, using jigdo-lite, and the main Debian Archive mirror (http://archive.debian.org/debian/) and jigdo indexes from https://cdimage.debian.org/mirror/cdimage/archive/3.1_r8/source/jigdo-dvd/. Everything is fine EXCEPT three files in debian-31r8-source-2.iso: pool/main/z/zope-quotafolder/zope-quotafolder_0.1.1.orig.tar.gz pool/main/s/sysvbanner/sysvbanner_1.0.orig.tar.gz pool/main/c/cl-md5/cl-md5_1.8.5.orig.tar.gz I get the following message: Copied input files to temporary file `debian-31r8-source-2.iso.tmp' - repeat command and supply more files to continue 2024-12-12 22:02:56 URL: https://archive.debian.org/debian/pool/main/z/zope-quotafolder/zope-quotafolder_0.1.1.orig.tar.gz [12490/12490] -> "./debian-31r8-source-2.iso.tmpdir/ archive.debian.org/debian/pool/main/z/zope-quotafolder/zope-quotafolder_0.1.1.orig.tar.gz" [1] 2024-12-12 22:02:56 URL: https://archive.debian.org/debian/pool/main/s/sysvbanner/sysvbanner_1.0.orig.tar.gz [1781/1781] -> "./debian-31r8-source-2.iso.tmpdir/ archive.debian.org/debian/pool/main/s/sysvbanner/sysvbanner_1.0.orig.tar.gz" [1] 2024-12-12 22:02:56 URL: https://archive.debian.org/debian/pool/main/c/cl-md5/cl-md5_1.8.5.orig.tar.gz [9103/9103] -> "./debian-31r8-source-2.iso.tmpdir/ archive.debian.org/debian/pool/main/c/cl-md5/cl-md5_1.8.5.orig.tar.gz" [1] After trying a few other mirrors as suggested, I was able to get sysvbanner_1.0.orig.tar.gz from http://debian.ethz.ch/debian-archive/debian/. No luck with the other two. I guess the ETHZ mirror was not synced for some time, it has versions up to Stretch. Which means something happened to this file in the main archive. After a while, I checked on Internet Archive, and hooray, it has the source ISOs at https://ia801309.us.archive.org/33/items/debian_3.1r8_source/ After downloading, I can confirm the SHA512 checksums match the ones from the Jigdo index. All in all, I have these files: ISO from Internet Archive 9026 2004-12-04 18:47 cl-md5_1.8.5.orig.tar.gz 2268 2003-02-16 17:02 sysvbanner_1.0.orig.tar.gz 12698 2005-01-04 19:02 zope-quotafolder_0.1.1.orig.tar.gz Debian Archive 9103 2009-09-10 13:32 cl-md5_1.8.5.orig.tar.gz 1781 1997-02-26 07:51 sysvbanner_1.0.orig.tar.gz 12490 2011-11-01 01:48 zope-quotafolder_0.1.1.orig.tar.gz ETHZ 9103 2009-09-10 13:32 cl-md5_1.8.5.orig.tar.gz 2268 2005-04-30 18:36 sysvbanner_1.0.orig.tar.gz 12490 2011-11-01 01:48 zope-quotafolder_0.1.1.orig.tar.gz I checked from copies of Debian 4.0r8 source ISOs I made last year (on 2023-12-13, surprisingly). zope-quotafolder was removed, cl-md5_1.8.5.orig.tar.gz (from 3.1_r8) was replaced with the older cl-md5_1.8.orig.tar.gz, and sysvbanner_1.0.orig.tar.gz is the same as in the ISOs from Internet Archive. However, I can't remember if I had to change the mirror for this, so not very useful. I also checked from Ubuntu archives: http://old-releases.ubuntu.com/ubuntu/pool/universe/c/cl-md5/cl-md5_1.8.5.orig.tar.gz http://old-releases.ubuntu.com/ubuntu/pool/universe/s/sysvbanner/sysvbanner_1.0.orig.tar.gz http://old-releases.ubuntu.com/ubuntu/pool/universe/z/zope-quotafolder/zope-quotafolder_0.1.1.orig.tar.gz Both sysvbanner_1.0.orig.tar.gz and zope-quotafolder_0.1.1.orig.tar.gz are identical to the ISOs from Internet Archive, and cl-md5_1.8.5.orig.tar.gz is the same as in Debian Mirror. Now, checking the contents: * Both versions of cl-md5_1.8.5.orig.tar.gz have identical files. The archive seems to have been re-created later with exactly the same files. * Both versions of zope-quotafolder_0.1.1.orig.tar.gz have identical files. However, one (from Internet Archive) is a tar.gz inside a tar.gz, not the other. * Files from sysvbanner_1.0.orig.tar.gz are almost identical as well: in the file from Debian Archive it's a Makefile and a C source file. In the ISO from Internet Archive, it's a text file to be cut to extract the C file, and it's identical to the other one except some spacing and indentation. My question is thus: does anyone have any idea why and how this happened? Why did source files from Debian get modified, while keeping their filename (hence version), thus breaking old ISO archives? Is it maybe because this is a fairly old version, and the mirror mechanism wasn't foolproof? Note: it's all hobby stuff, nothing really important for me, I'm just trying to make sense of this. Best regards, Jean-Claude Arbaut
