Tails

#1 Updated by CyrilBrulebois 2020-04-07 06:38:01

For 4.5 specifically, given 4.5~rc1 is the release introducing big bootloader-side changes plus the aufs→overlayfs switch, and given RCs are likely less tested than regular releases, I decided to test many combinations (even if intrigeri explained very well that many changes should just work and why — and it was very convincing). Part of those scenarios were already tested when releasing 4.5~rc1, but I wanted to ensure we tested what happened with an upgrade in two steps, to see if upgrading to 4.5~rc1 could possibly lead to issues for the next release.

Everything looks good, very well done, intrigeri & segfault!

In the following, on my Thinkpad X230:

• UEFI := UEFI Only, CSM disabled, Secure Boot disabled
• UEFI+SB := UEFI Only, Secure Boot enabled
• I didn’t test anything allowing UEFI+BIOS at the same time.

In short, I tested (boot-to-greeter-or-desktop, not deeper):

• 4.4.1 → 4.5 (test channel): all possible combinations, meaning {initial boot, after-upgrade boot} = {BIOS, UEFI} × {BIOS, UEFI, UEFI+SB}.
• 4.4.1 → 4.5~rc1 (alpha channel) → 4.5 (test channel) in 3 ways:
  • BIOS → BIOS → {BIOS, UEFI, UEFI+SB}: aka. “using BIOS until now, switching to UEFI w/ or w/o SB in the end”
  • UEFI → UEFI → {UEFI, UEFI+SB, BIOS}: aka. “using UEFI already, switching to UEFI+SB in the end, while still being able to use non-fancy boots”
  • UEFI → UEFI+SB → {UEFI+SB, UEFI, BIOS}: aka. “using UEFI already, switching to UEFI+SB in the middle, while still being able to use non-fancy boots in the end”

I haven’t tested direct installs of 4.5~rc1 then upgrade to 4.5, which might make sense because that would be the first time we upgrade systems initially installed with Secure Boot and overlayfs. I do not consider these tests time-critical though, as users having bootstrapped material with a release candidate might be expected to perform some fixups if that was necessary, and we wouldn’t block the 4.5 release on this. I’ll try and get that tested once I’m done with my other manual tests.

For completeness, my notes follow:

UEFI boot := UEFI Only / CSM: disabled.


4.4.1: direct to 4.5
------
BIOS boot → upgrade to 4.5.   OK
Still boots under BIOS.       OK
Also boots under UEFI.        OK
Also boots under UEFI+SB.     OK

4.4.1: direct to 4.5
------
UEFI boot → upgrade to 4.5.   OK
Still boots under UEFI.       OK
Also boots under UEFI+SB.     OK
Also boots under BIOS.        OK


4.4.1: via 4.5~rc1 to 4.5
------
UEFI boot → upgrade to 4.5~rc1.      OK
Still boots under UEFI.              OK
UEFI boot → upgrade to 4.5.          OK
Still boots under UEFI.              OK
Also boots under UEFI+SB.            OK
Also boots under BIOS.               OK

4.4.1: via 4.5~rc1 to 4.5
------
BIOS boot → upgrade to 4.5~rc1.      OK
Still boots under BIOS.              OK
BIOS boot → upgrade to 4.5.          OK
Still boots under BIOS.              OK
Also boots under UEFI.               OK
Also boots under UEFI+SB.            OK

4.4.1: via 4.5~rc1 to 4.5
-----
UEFI boot → upgrade to 4.5~rc1.      OK
Still boots under UEFI.              OK
Also boots under UEFI+SB.            OK
UEFI+SB boot → upgrade to 4.5.       OK
Still boots under UEFI+SB.           OK
Also boots under UEFI.               OK
Also boots under BIOS.               OK


NOT TESTED: 4.5~rc1 from scratch to 4.5.

#3 Updated by CyrilBrulebois 2020-04-18 01:00:30

Also tested: 4.5~rc1 initially deployed with Secure Boot enabled (the last setup I didn’t test yet) → 4.5, as part of a check for my Bug #17614 manual fix, and that worked fine; still booting after upgrade to 4.5 with UEFI+SB, UEFI, and BIOS.

As far as I’m concerned, all tests look very good; we only need to tweak the doc a little as agreed in the previous comment.