Switching to to xz -0 from pxz increases compressed raspi2 image size

Compr. Threads Wall.t User.t Sys.t Res.t Wall.t User.t Sys.t Res.t Size(kB)
-0 -T1 77.68 68.78 1.10 4916 25.04 23.07 0.14 2516 365040
-1 -T1 100.05 89.72 1.24 11020 21.79 20.07 0.09 3360 322660
-2 -T1 127.23 118.40 1.31 18760 20.88 19.12 0.08 4292 309644
-3 -T1 165.56 160.31 1.30 34060 20.28 18.62 0.10 6388 302344
-4 -T1 257.81 252.85 1.26 50688 19.54 18.20 0.07 6364 284368
-5 -T1 322.24 318.51 1.41 97724 19.51 17.95 0.09 10576 272192
-6 -T1 374.25 366.92 1.39 97568 19.28 17.92 0.11 10664 270496
-7 -T1 389.19 385.40 1.44 191940 18.11 17.61 0.14 18664 266972
-8 -T1 405.00 400.37 1.73 380276 17.41 17.01 0.10 35028 258024
-9 -T1 486.82 436.84 3.08 690076 16.81 16.24 0.13 67888 251436
-0 -T4 37.83 93.87 1.32 25000 24.91 23.35 0.10 2652 368880
-1 -T4 45.74 138.40 1.59 72108 22.46 20.58 0.10 3304 330744
-2 -T4 56.26 175.05 1.86 126880 20.49 19.52 0.10 4440 315876
-3 -T4 68.22 240.02 1.93 235104 20.29 18.84 0.16 6360 307180
-4 -T4 94.85 348.37 2.52 300212 20.71 19.62 0.11 6488 289376
-5 -T4 118.08 435.77 2.88 556820 19.67 18.34 0.13 10580 276576
-6 -T4 133.11 495.87 2.37 556876 18.85 18.30 0.12 10468 274912
...

With -T4 -7 the Bionic VM with 1GB RAM which I tested the combinations in started swapping but probably we would not want to pick a better compression. If so I'll add the missing results.

-6 is the default, why would we be using -0?

> Looking at the Xenial numbers I believe -3 would result reasonably fast
> builds with a tolerable increase in size compared to xz -6's results.

As you've seen, the rpi2 server image is the only place we're using this code today. We are not overly concerned with the build speed as these are highly-parallelizable builds that are not part of a critical pipeline. I believe we should go with xz -6; certainly for bionic and forward where we have actual parallelization, the build time penalty appears to be negligible.

In development branch xz -0 is used for compressing the Ubuntu Core image as well but xz's compression setting have little impact on its compressed size thus -0 can be kept there:

(ubuntu-core-16-amd64.img)
ubuntu@bb:~/test-core$ cat time-size.txt
Compr. Threads Wall.t User.t Sys.t Res.t Wall.t User.t Sys.t Res.t Size(kB)
-0 -T1 79.97 76.04 1.54 5068 10.37 6.82 0.08 2528 381472
-1 -T1 119.13 115.42 1.76 10952 11.65 6.90 0.10 3376 379540
-2 -T1 148.43 142.16 1.74 18612 11.83 6.92 0.11 4440 379504
-3 -T1 164.89 158.90 1.93 34076 10.05 6.88 0.14 6416 379496
-4 -T1 178.44 172.44 1.85 50616 11.58 6.91 0.13 6296 379464
-5 -T1 187.43 181.44 2.13 97644 17.77 7.21 0.10 10484 379444
-6 -T1 201.22 194.96 2.03 97716 14.27 8.59 0.15 10572 379444
-7 -T1 201.56 194.90 1.97 191952 14.57 8.53 0.15 18660 379444
-8 -T1 200.86 195.67 2.10 380228 13.11 8.61 0.10 35112 379440
-9 -T1 283.20 238.30 4.75 689920 16.62 8.59 0.13 68004 379440
-0 -T4 33.40 109.76 2.43 24920 12.12 8.47 0.09 2744 381844
-1 -T4 56.35 174.24 2.91 73592 11.99 8.51 0.10 3296 379668
-2 -T4 55.58 185.15 3.82 141184 12.63 8.55 0.10 4376 379572
-3 -T4 55.92 194.41 3.15 276356 11.82 8.67 0.10 6348 379528
-4 -T4 63.59 218.39 3.29 342144 11.74 8.57 0.09 6416 379496
-5 -T4 64.24 226.16 5.84 675728 12.31 8.56 0.12 10488 379460
-6 -T4 67.20 239.20 5.55 675512 12.88 8.54 0.15 10652 379456
-7 -T4 3539.73 306.44 98.61 740100 12.55 8.60 0.12 18632 379448
Command terminated by signal 2
-8 -T4 34404.61 126.70 410.87 720576

This bug was fixed in the package livecd-rootfs - 2.485

---------------
livecd-rootfs (2.485) bionic; urgency=medium

  [ Michael Hudson-Doyle ]
  * Automount a partition of fixed UUID for giving subiquity canned answers.

  [ Balint Reczey ]
  * Use default xz compression (-6) for raspi2 image.
    Xz with -0 generated ~35% bigger compressed image. (LP: #1732681)
  * Zero fill space in UEFI partitions, too
  * Set LB_ISO_PREPARER to livecd-rootfs to avoid unbound variable in
    default string
  * Build minimized images in a 1GiB image

 -- Michael Hudson-Doyle <email address hidden> Mon, 04 Dec 2017 10:51:13 +1300

Since the change passing -0 to xz never entered Xenial on its own this bug does not have to be tracked for Xenial.