copy/multiple: split selection of images to be copied in copyMultipleImages

[flouthoc]

Following PR splits copyMultipleImages into a new function called
prepareCopyInstances which returns a list of instanceCopy.

prepareCopyInstances does that in these two steps

• How to copy via instanceCopyKind
• What to copy by the content of the returned list.

See newly added unit test to see how it is used.

[flouthoc]

@mtrmac PTAL

[mtrmac]

This works well enough right now, but it will ~immediately need to be replaced when implementing the “add a new Zstd variant” feature; there’s no clean way to represent that.

We could hack that with ListOperation: ListOpAdd, AddDigest: theOldDigestToConvert; that would be ugly, but it would work once for the “add a Zstd variant” feature. But then someone may want an “add a gzip variant”… and we are missing a field where to recordwhat kind of modification we should make.

So I think it’s cleaner to add a separate struct immediately, like the original version of #1883 did e.g. in 941cd13 .

[flouthoc]

@mtrmac How about new struct could look like this ?

type addOperation int

const (
	addop_compress_gzip addOperation = iota
	addop_compress_zstd
)

type instanceCopy struct {
        instance internal.ListEdit
        addOp addOperation
}

[mtrmac]

• That structure wouldn’t allow a non-converting copy that just copies layers as is
• I don’t know what prepareEditList would fill in for most of the fields of instance.

---

I was thinking vaguely

/* enum instanceCopyKind {
   instanceCopyCopy // a single copy, replacing the original
   instanceCopyClone // creating a new copy of the original, not replacing it
} */

type instanceCopy {
   op instanceCopyKind
   sourceDigest // digest in the source == instance ID
   // otherCopyOptions: for now maybe just
   forceZstd bool
}

That’s assuming that after we do an instance copy, we have some other code to determine that we need to adjust the Zstd annotation, and that the annotation update doesn’t need to be represented here.

Maybe we don’t need a new instanceCopyKind enum (for now? at all?), just a clone bool (or createCloneInstance?) would be sufficient.

[mtrmac]

Thanks! A few more cleanups, please.

[flouthoc]

@mtrmac PTAL again :)

[mtrmac]

Thanks! LGTM.

[flouthoc]

@mtrmac @giuseppe @rhatdan PTAL