fixes when no equal sign in label file

[lifubang]

Signed-off-by: Lifubang lifubang@acmcoder.com

- What I did
When use --label:
./build/docker run -d --name=redis5 --label foo redis

root@dockerdemo:~/gocode/src/github.com/docker/cli# ./build/docker inspect --format '{{.Config.Labels}}' redis5
map[foo:]

When use --label-file:

root@dockerdemo:~/gocode/src/github.com/docker/cli# echo $foo
bar
root@dockerdemo:~/gocode/src/github.com/docker/cli# cat ~/labels.txt 
foo
bar=
root@dockerdemo:~/gocode/src/github.com/docker/cli# build/docker run -d --name=redis7 --label-file ~/labels.txt redis
root@dockerdemo:~/gocode/src/github.com/docker/cli# ./build/docker inspect --format '{{.Config.Labels}}' redis7
map[bar:]

label foo is lost.

The behavior of --label and --label-file is different when there is no equal sign after label variable.
- How I did it
When there is no equal sign after label variable, and no emptyFn, add it to output.

- How to verify it
After fix:

root@dockerdemo:~/gocode/src/github.com/docker/cli# build/docker run -d --name=redis8 --label-file ~/labels.txt redis
root@dockerdemo:~/gocode/src/github.com/docker/cli# ./build/docker inspect --format '{{.Config.Labels}}' redis8
map[bar: foo:]

- Description for the changelog
fixes when no equal sign in label file

- A picture of a cute animal (not mandatory but encouraged)

[codecov-io]

Codecov Report

Merging #1416 into master will increase coverage by 0.13%.
The diff coverage is 50%.

@@            Coverage Diff             @@
##           master    #1416      +/-   ##
==========================================
+ Coverage   54.26%   54.39%   +0.13%     
==========================================
  Files         289      289              
  Lines       19331    19268      -63     
==========================================
- Hits        10490    10481       -9     
+ Misses       8165     8115      -50     
+ Partials      676      672       -4

[vdemeester]

LGTM 🐸

[thaJeztah]

Haven't checked yet, but is this same function also used for env-files? If so, should it behave the same?

We should at least have a test for this, as the behavior of empty values vs not-set is already tricky, so easy to introduce regressions / change in behavior

[vdemeester]

Haven't checked yet, but is this same function also used for env-files? If so, should it behave the same?

Yeah, but for env-files the emptyfn is not nil so it should still act the same 😉 But I agree on a test

[lifubang]

Thank you for all your review, I'll add some test cases.

[lifubang]

I have add two test cases for label/env file without equal sign.

[lifubang]

@thaJeztah I don't know if I have understood your points. Do you mean add some test cases, or change code to check the lines value?

[tonistiigi]

ping @thaJeztah

[thaJeztah]

Cleaning up old PRs; I thought this was fixed, but indeed looks like the "no value set" omits the label; still need to look if that was or was not by design, but did a quick rebase, and updated the patch to make the new tests added in this PR pass.

[codecov-commenter]

Codecov Report

❌ Patch coverage is 0% with 3 lines in your changes missing coverage. Please review.

Files with missing lines	Patch %	Lines
opts/parse.go	0.00%	3 Missing ⚠️

📢 Thoughts on this report? Let us know!