feat(26.04): add e2fsprogs

[Meulengracht]

Proposed changes

The e2fsprogs suite is needed by Ubuntu Core 26. This needs logsave (#879) and systemd changes (#878)

Checklist

• I have read the contributing guidelines
• I have tested my changes (see how)
• I have already submitted the CLA form

[lczyk]

• blocked on ci error preparing lxd machine in test #828. working on a fix. for now cancelled the integration tests as they would never succeed and eventually time out, and they were taking up worker allocation. will re-trigger when ci is back to a working state. apologies for the inconvenience!

[lczyk]

• ci error preparing lxd machine in test #828 resolved. removing 'blocked' and updating with the base branch

[alesancor1]

Blocked by #879

[alesancor1]

I understand that standard contains everything, but we should test each of the slices separately to guarantee that they can work standalone.

For a better coverage I'd suggest to just run some smoke tests for the standard slice (just run -v and -h for the binaries) and move the functionality testing to a test for each individual slice.

Use variant and separated sh scripts for each variant (anticipating @lczyk's comment :p ) - For reference: example

[lczyk]

yep. seconded ☝️ :) makes for a muuuch cleaner code imo (bash linting + syntax highlight of the test files) + spread runs all the bits as separate tests, so if one breaks it does not take the whole test down

[alesancor1]

Not a PR specific comment.

@lczyk @cjdcordeiro we have a bunch of tests for libraries like this one, but some other libraries are not tested - should we put a convention in place to always do a simple existence check for the files?

[lczyk]

i think lots of lower hanging fruit and i'm not convinced that simple existence test is really testing anything that installability tests are not testing already. honestly, i'd even say to remove this and libss2 tests

[lczyk]

• feat(26.04): add logsave #879 is merged so marking as unblocked and updating from trunk

[lczyk]

yep. seconded ☝️ :) makes for a muuuch cleaner code imo (bash linting + syntax highlight of the test files) + spread runs all the bits as separate tests, so if one breaks it does not take the whole test down

[lczyk]

i think lots of lower hanging fruit and i'm not convinced that simple existence test is really testing anything that installability tests are not testing already. honestly, i'd even say to remove this and libss2 tests

[lczyk]

Suggested change

	- libc6_libs
	- libc6_libs
	- libcom-err2_libs

libcom-err2_libs requires libcom_err.so.2 from libcom-err2. here's ldd output:

	linux-vdso.so.1 (0x0000ffff8084c000)
	libcom_err.so.2 => /usr/lib/aarch64-linux-gnu/libcom_err.so.2 (0x0000ffff80740000)
	libc.so.6 => /usr/lib/aarch64-linux-gnu/libc.so.6 (0x0000ffff80540000)
	/lib/ld-linux-aarch64.so.1 (0x0000ffff80810000)

[lczyk]

Suggested change

	/usr/sbin/fsck.ext2:
	/usr/sbin/fsck.ext3:
	/usr/sbin/fsck.ext4:
	/usr/sbin/fsck.ext2: # Symlink to e2fsck
	/usr/sbin/fsck.ext3: # Symlink to e2fsck
	/usr/sbin/fsck.ext4: # Symlink to e2fsck

[lczyk]

Suggested change

	/usr/sbin/mkfs.ext2:
	/usr/sbin/mkfs.ext3:
	/usr/sbin/mkfs.ext4:
	/usr/sbin/mkfs.ext2: # Symlink to mke2fs
	/usr/sbin/mkfs.ext3: # Symlink to mke2fs
	/usr/sbin/mkfs.ext4: # Symlink to mke2fs

[lczyk]

e2scrub and e2scrub_all are bash scripts. they should be in a slice called scripts, and have a dependency on bash_bins, base-files_bin (due to the shebang) and a pile of utils. then, if you want a slice named after a function, you could do something like this:

scrub:
  essential:
    - e2fsprogs_scripts

scripts:
  ...

[lczyk]

also these should be tested. it looks like they both have help if you supply some non-existent flag. you can then assert the return code is 2. this is a bit annoying to do with the -e flag (which i nonetheless recommend strongly. here is a small working example

#!/usr/bin/env -S bash -ex

function my_program {
  echo "doing something"
  return 33
}

status_file="$(mktemp)"
trap "rm -f $status_file" EXIT

out="$(
  set +e
  my_program
  echo "$?" >"$status_file"
)"
code="$(cat "$status_file")"

echo "code: $code"
echo "output: $out"

[lczyk]

also, i've done the analysis of e2scrub and e2scrub_all and they depend on the following binaries:

e2scrub:

echo
awk
date
dumpe2fs
fstrim
grep
lsblk
lvs
realpath
sleep

e2scrub_all:

echo
dumpe2fs
grep
lsblk
lvs
readlink
sed
systemd-escape

i think you can skip the systemd part since the script accounts for it possibly not being there, but do add a comment saying that systemd-escape is needed for full systemd functionality

[lczyk]

Suggested change

	/usr/sbin/e2label:
	/usr/sbin/e2label: # Symlink to tune2fs

[lczyk]

Suggested change

	/usr/sbin/e2mmpstatus:
	/usr/sbin/e2mmpstatus: # Symlink to dumpe2fs

[lczyk]

• adding 'blocked' since blocked on feat(26.04): improve systemd slices #878