Bug#592519: initramfs-tools: COMPRESS option should be more flexible

[Rogério Brito <rbrito@ime.usp.br>]

Package: initramfs-tools
Version: 0.139
Followup-For: Bug #592519
X-Debbugs-Cc: benh@debian.org, mika@debian.org, debian-kernel@lists.debian.org


Hi.

In a similar vein to what was reported by Roger Shimizu, it takes ages to
compress with xz on an armel system with only 128MB of RAM. By default,
(which is what is used) xz uses level 6 compression and that is what takes a
lot of time to compress.

I switched to gzip and the initrd is quite larger. I just performed some
tests with perf stat running 3 times a lot of commands to compress my
current initrd.img-5.8.0-1-marvell and I found xz -0 both to be faster than
gzip -9 *and* generating a smaller file as a result.

Here are some results that I gathered. I didn't wait for xz with the default
-6 level to finish...

| compressor           |   time |     size |
|----------------------+--------+----------|
| gzip -9nk            | 69.264 |  5980207 |
| gzip -nk             | 34.468 |  5995619 |
| xz -0k               | 35.280 |  5000752 |
| xz -1k               | 45.351 |  4868388 |
| xz --check=crc32 -1k | 44.902 |  4868384 |
| bzip2 -k             | 51.750 |  5709152 |
| uncompressed         |      0 | 12965376 |

Passing an option to xz (or, in general, to any compressor) would be,
indeed, very nice.

OTOH, on a fast machine, I would like to use something like xz -9e (that is,
maximum compression---probably overkill for not so large files---but with
"extreme" searches for more compression).

Thanks,

Rogério Brito.

-- 
Rogério Brito : rbrito@{ime.usp.br,gmail.com} : GPG key 4096R/BCFCAAAA
http://cynic.cc/blog/ : github.com/rbrito : profiles.google.com/rbrito
DebianQA: http://qa.debian.org/developer.php?login=rbrito%40ime.usp.br