Let graphdrivers declare diff stream fidelity#32092
Conversation
|
Design ok for me. |
There was a problem hiding this comment.
Maybe add an extra doc comment like
In practice when true, relies on the graphdriver to provide consistent tar streams for a layer rather than the generic algorithm the layer store provides
I don't usually like stuff like this that refers to "parent" packages, but is helpful for readers at least.
|
Design LGTM. |
|
updated the doc comment; see if that reads ok for you @cpuguy83 |
|
Perfect! |
|
The single test failure looks unrelated, and I see other PR's that have it as well. |
cpuguy83
left a comment
cpuguy83
left a comment
There was a problem hiding this comment.
I wonder if we can refactor this a bit to hide the complexity behind an interface rather than storing graphdriver caps in the layer store.
|
how about if the layer store just has a |
|
I made that change, and I think that reads a little better: at init time, the layer store initializes itself based on the driver's capabilities. |
|
from offline conversation with @cpuguy83 : should also update the graphdriver plugin docs |
This allows graphdrivers to declare that they can reproduce the original diff stream for a layer. If they do so, the layer store will not use tar-split processing, but will still verify the digest on layer export. This makes it easier to experiment with non-default diff formats. Signed-off-by: Alfred Landrum <[email protected]>
|
ping @tiborvass - @cpuguy83 suggested you for a second code review |
|
ping @thaJeztah - anything else I can do to push this over the line? |
|
LGTM |
9 participants
This allows graphdrivers to declare that they can reproduce the original
diff stream for a layer. If they do so, the layer store will not use
tar-split processing, but will still verify the digest on layer export.
This makes it easier to experiment with non-default diff formats.
Signed-off-by: Alfred Landrum [email protected]