Add the ability for images to retain history if provided #202
Conversation
remote/save.go
Outdated
| if len(cfg.History) != len(layers) { | ||
| cfg.History = make([]v1.History, len(layers)) | ||
| } |
There was a problem hiding this comment.
We shouldn't need this as we've been careful to update the history when instantiating an image and adding layers, but you never know...
natalieparellano
force-pushed
the
retain-history
branch
from
463e531
to
3822f51
Compare
Signed-off-by: Natalie Arellano <narellano@vmware.com>
natalieparellano
force-pushed
the
retain-history
branch
3 times, most recently
from
3f20ac4
to
97c0a45
Compare
Signed-off-by: Natalie Arellano <narellano@vmware.com>
Signed-off-by: Natalie Arellano <narellano@vmware.com>
Signed-off-by: Natalie Arellano <narellano@vmware.com>
natalieparellano
force-pushed
the
retain-history
branch
from
97c0a45
to
275a68c
Compare
Pointers to interfaces are weird Signed-off-by: Natalie Arellano <narellano@vmware.com>
Signed-off-by: Natalie Arellano <narellano@vmware.com>
Signed-off-by: Natalie Arellano <narellano@vmware.com>
|
This should be done except for updating the fakes. I'm going to try to pull this into the lifecycle and then I will update the fakes with whatever is needed. In the meantime this should be ready for some initial feedback. |
Signed-off-by: Natalie Arellano <narellano@vmware.com>
image.go
Outdated
| if err != nil { | ||
| return nil, err | ||
| } | ||
| retImage := image |
There was a problem hiding this comment.
ok this is maybe my own confusion(s) here. but v1.Image is (probably) an interface, right? which means here you're doing a pointer copy, right? so below when you call mutate on the retImage doesn't that also change the image param that was passed in? (that would be a reason to do a deepcopy here, imo, if you're trying not to change your input param)
There was a problem hiding this comment.
Summarizing offline discussion - I think we agreed that this is weird, but it is the way to do it unless you want to use a pointer to interface, which is also weird
…er explicitly asked for history when the image was constructed Signed-off-by: Natalie Arellano <narellano@vmware.com>
natalieparellano
force-pushed
the
retain-history
branch
from
d47a25c
to
e99e711
Compare
Signed-off-by: Natalie Arellano <narellano@vmware.com>
…quested Signed-off-by: Natalie Arellano <narellano@vmware.com>
| additions := make([]mutate.Addendum, 0) | ||
| for _, layer := range layers { | ||
| if len(history) != len(layers) { | ||
| history = make([]v1.History, len(layers)) |
There was a problem hiding this comment.
Will this lead to an off by N type of association if we get here with partial history? Not super worried about that but curious.
There was a problem hiding this comment.
I think we should be okay... we "normalize" the history (remove empty layers, check that the lengths are equal) at every stage - on initialization, when adding a layer, on save.
We should fetch the config file AFTER we've normalized history, to avoid writing invalid history Signed-off-by: Natalie Arellano <narellano@vmware.com>
This is a first step toward buildpacks/lifecycle#411
I think this does more or less what is expected, but I need to add tests 😈