Handle default bootloader entry patterns correctly

[benmaddison]

Resolves #183

This update implements matching of boot entry names against the pattern in the
default option from loader/loader.conf, intended to be compatible with the
behaviour of systemd-boot.

Outstanding work

• implement [..] style patterns - deferred. Tracked in support for [..] patterns in loader default entry patterns #191
• better error handling, fewer panics
• discuss use of unstable slice_take feature

[rbradford]

@benmaddison Looks like a great start. I'm very happy with what you've done so far.

Some minor things:

1. Could you split the comment updates/reformatting into a separate commit
2. Try and follow the existing commit message style (with the starting "fat:" etc..)
3. Don't forget your "-s"

On the "slice_take" thing. I don't know ... do you think it's on a path for staying around or might be dropped...it seems quite a stale nightly feature with not much activity.

I would also prefer to fix the broken OS support right now than perhaps addressing missing feature gaps unless they are in active use (that could be in a follow-up!)

[benmaddison]

Some minor things:

1. Could you split the comment updates/reformatting into a separate commit

2. Try and follow the existing commit message style (with the starting "fat:" etc..)

3. Don't forget your "-s"

Yup - I plan to rebase and squash before asking for a final review

On the "slice_take" thing. I don't know ... do you think it's on a path for staying around or might be dropped...it seems quite a stale nightly feature with not much activity.

I'm not sure that I know the answer to that. I'll take another look at it to see if I can re-implement that loop without loosing too much readability.

I would also prefer to fix the broken OS support right now than perhaps addressing missing feature gaps unless they are in active use (that could be in a follow-up!)

Ack. Agreed.

[benmaddison]

Open issues for review:

• I have kept the use of slice_take: I'm not seeing a way of avoiding it without effectively re-implementing that feature here. I'm happy to commit to doing the re-factor in future if the feature is dropped or this project wants to move to stable before the feature is stabilized.
• find_entry will currently bail out with an error if a call to compare_entry fails (due to an unterminated string), rather than falling back to a valid entry that was found earlier. Making that happen is not hard, but will be verbose, and I'm not sure it's worth the bother since that can only really happen if the FAT FS is broken. Guidance welcome.

[rbradford]

I would love to test with this an integration test? Is there a cloud image I can use?

Open issues for review:

* I have kept the use of `slice_take`: I'm not seeing a way of avoiding it without effectively re-implementing that feature here. I'm happy to commit to doing the re-factor in future if the feature is dropped or this project wants to move to stable before the feature is stabilized.

👍

* `find_entry` will currently bail out with an error if a call to `compare_entry` fails (due to an unterminated string), rather than falling back to a valid entry that was found earlier. Making that happen is not hard, but will be verbose, and I'm not sure it's worth the bother since that can only really happen if the FAT FS is broken. Guidance welcome.

👍 I don't think we need to worry about dealing with a corrupt FAT.

[rbradford]

Really happy with this. Just some very minor nits.

[benmaddison]

I have group_imports = StdExternalCrate in my local fallback rustfmt.toml file, which was why that grouping was occurring.

To ensure that doesn't happen to future contributors (incl. me), I have dropped an empty rustfmt.toml into the repo root, so that the defaults get used.

[benmaddison]

I would love to test with this an integration test? Is there a cloud image I can use?

The NixOS image that I was using when I found the issue is locally generated, and the downloadable ones ship with grub.

I've had a search around for other options. So far I've checked ubuntu, fedora and arch: they all use grub :-(

I think the best I can offer is to generate a minimal nixos one, and store it in the repo with LFS, but it will make cloning unpleasant.

[rbradford]

I have group_imports = StdExternalCrate in my local fallback rustfmt.toml file, which was why that grouping was occurring.

To ensure that doesn't happen to future contributors (incl. me), I have dropped an empty rustfmt.toml into the repo root, so that the defaults get used.

Ah, cool.