Conversation
|
Thanks Tibor! |
|
I don't like separate file for every action. Also, go to definition doesn't work with jobs, so I need to inspect all |
|
@LK4D4 separate file for every action is so much nicer than one file with 2k lines |
|
These PRs are hard to review so can we trust you that this did not change functionally in an external way and that it is just refactoring and moving code around? |
|
What makes them hard to review specifically? Anything I can do to fix that? |
|
@shykes no, it's just the big blobs of green in the github UI. Nothing you guys did wrong. ;) |
|
@LK4D4 it doesn't have to stay 1-file-per-action forever, but for this particular operation (transcribing large amounts of code across directories) it definitely helps. |
|
@crosbymichael @shykes I could rearrange this adding one commit before all those, which would be without moving the blocks of codes, so that you can see a clean diff in the same file (but it would not compile), and then have each of those "Move" commits be a trivial cut/paste. |
|
@tiborvass no it's ok. |
|
@crosbymichael the methodology was to change as little as possible in the actual content. In most cases that was:
I will try to point out exceptions to this in line comments. |
|
Just worked with splitted to 40 files daemon. This was awful. Such code arranging will slow my development I think. I agreed that 2k lines is too much, but 40 files with 20-40 lines of code is other extreme which is not better. I think 600-1000 lines is still okay and there are many such files in |
graph/service.go
Outdated
There was a problem hiding this comment.
Because we are not checking errors returned from register we do not know if this refactoring is overwriting any jobs already registered with the same name throughout the code base.
Do you think we can check the errors from register?
There was a problem hiding this comment.
In that case I recommend going through a loop a-la daemon.Daemon.Install
|
@LK4D4 really I'm not saying we should switch everything to lots of little files. It just seemed right to better manage a big copy-paste operation with potential merge conflicts etc. |
|
@crosbymichael @shykes updated. |
|
Travis is failing now |
Signed-off-by: Solomon Hykes <solomon@docker.com>
Signed-off-by: Solomon Hykes <solomon@docker.com>
Signed-off-by: Solomon Hykes <solomon@docker.com>
Signed-off-by: Solomon Hykes <solomon@docker.com>
Signed-off-by: Solomon Hykes <solomon@docker.com>
Signed-off-by: Solomon Hykes <solomon@docker.com>
Note: this cannot yet be moved to graph/ because of a lingering dependency on daemon. This has been noted in a FIXME. Signed-off-by: Solomon Hykes <solomon@docker.com>
Note: these 2 jobs should be merged into one. This was noted in a FIXME. Signed-off-by: Solomon Hykes <solomon@docker.com>
Signed-off-by: Tibor Vass <teabee89@gmail.com>
Signed-off-by: Tibor Vass <teabee89@gmail.com>
|
@crosbymichael sorry about that, it should be good now! |
|
LGTM |
|
Ping @unclejack |
|
Sorry wrong pr, ignore last message. |
|
Closing in favor of #7452 |
This PR removes all the image jobs from server, into graph/. There is one exception:
image_deletewhich still has a dependency on the daemon (to be fixed); that has been moved to daemon/.All this work was done by @shykes. I simply resolved rebase conflicts and migrated tests.
Ping @crosbymichael @LK4D4 @unclejack @erikh