Bug #12739

Metadata for directories inside the squashfs not reproducible in some environments

Added by anonym 2017-06-19 10:17:54 . Updated 2017-10-02 12:00:41 .

Status:
Rejected
Priority:
Normal
Assignee:
Category:
Build system
Target version:
Start date:
2017-06-19
Due date:
% Done:

0%

Feature Branch:
Type of work:
Research
Blueprint:

Starter:
Affected tool:
Deliverable for:
289

Description

See Feature #12608#note-18 for arnaud’s diffoscope report. There are also .torrent:s for the good ISO, and for arnaud’s.

The diffs all look like this:

│ │  -rwxr-xr-x root/root              3499 2016-11-13 23:23 /usr/lib/gnupg1/gpgkeys_mailto
│ │ -drwxr-xr-x root/root                90 2017-05-21 18:33 /usr/lib/gnupg2
│ │ +drwxr-xr-x root/root               102 2017-05-21 18:33 /usr/lib/gnupg2
│ │  lrwxrwxrwx root/root                30 2017-02-14 00:29 /usr/lib/gnupg2/gpg-preset-passphrase -> ../gnupg/gpg-preset-passphrase

Subtasks


History

#1 Updated by anonym 2017-06-19 10:24:34

  • Assignee set to lamby
  • QA Check set to Info Needed

Could you have a look?

#2 Updated by intrigeri 2017-06-22 13:49:35

#3 Updated by lamby 2017-06-24 12:05:13

Two unconfirmed hypothesis at the moment:

a) This is just fallout from the existing “bleed” — https://labs.riseup.net/code/issues/12736

b) The hash table used to do directory lookups is non-deterministic — http://sources.debian.net/src/squashfs-tools/1:4.3-4/squashfs-tools/mksquashfs.c/#L1793

I recommend we fix all the bleed-related issue, then return to this and see whather it still exists.

#4 Updated by intrigeri 2017-06-24 17:17:14

> b) The hash table used to do directory lookups is non-deterministic — http://sources.debian.net/src/squashfs-tools/1:4.3-4/squashfs-tools/mksquashfs.c/#L1793

Note that we use lynxis’ version of mksquashfs: https://deb.tails.boum.org/pool/main/s/squashfs-tools/ (the version currently in Debian can’t produce reproducible SquashFS; I got in touch with the Debian maintainer, suggesting he switched to that new upstream, and got no reply so far; I’ll ping once lynxis has tagged & released a new version).
This might might, or might not, affect this hypothesis (I didn’t check). But for now, whatever, because:

> I recommend we fix all the bleed-related issue, then return to this and see whather it still exists.

Agreed.

#5 Updated by lamby 2017-06-25 08:48:23

intrigeri wrote:
> Note that we use lynxis’ version of mksquashfs: https://deb.tails.boum.org/pool/main/s/squashfs-tools/

Nod. Of course, lnyxis’ version could also not address this (potential/hypothetical) hash table issue yet. :)

> I got in touch with the Debian maintainer, suggesting he switched to that
> new upstream

Could it do with a new maintainer? :) Feel free to suggest that.

#6 Updated by intrigeri 2017-06-25 09:33:10

>> I got in touch with the Debian maintainer, suggesting he switched to that new upstream

> Could it do with a new maintainer? :) Feel free to suggest that.

Laszlo did some work on the package this month so I’m hopeful that once the upstream situation is clarified, we should be good on the Debian side :)

#7 Updated by intrigeri 2017-07-18 12:58:38

  • Target version changed from Tails_3.1 to Tails_3.2

(As discussed above, let’s fix known causes of “bleed” first.)

#8 Updated by anonym 2017-08-17 12:05:08

intrigeri wrote:
> (As discussed above, let’s fix known causes of “bleed” first.)

FYI, AFAICT the causes of “bleed” were Bug #12736 and Bug #12737, and they are both closed.

#9 Updated by intrigeri 2017-08-24 13:18:46

  • Assignee changed from lamby to anonym
  • QA Check deleted (Info Needed)

#10 Updated by intrigeri 2017-09-07 12:38:01

  • Status changed from Confirmed to Rejected

Apparently not an issue in itself: not seen it since the root causes were fixed.

#11 Updated by anonym 2017-09-28 18:55:56

  • Target version changed from Tails_3.2 to Tails_3.3

#12 Updated by anonym 2017-09-28 19:00:10

  • Target version changed from Tails_3.3 to Tails_3.2

#13 Updated by anonym 2017-10-02 12:00:42

  • Assignee deleted (anonym)