Stop using the "name" argument in the "install-plugin" CLI command

[jglick]

Cleans up a longstanding UX regression from #2795: there is no good reason to require the -name option to the install-plugin CLI command. At best it is a nuisance, and at worst it could mean data corruption from typos. So instead of

java -jar …/jenkins-cli.jar -s http://jenkins/ install-plugin -name some-thing = < target/some-thing.hpi

you can just write

java -jar …/jenkins-cli.jar -s http://jenkins/ install-plugin = < target/some-thing.hpi

since the manifest header

Short-Name: some-thing

already conveys this information.

Ditto for URL-based installation; the former code tried to parse out the short name from the URL path, which was pretty fragile. Now you can just

java -jar …/jenkins-cli.jar -s http://jenkins/ install-plugin http://download-server/this-naming-is-irrelevant.hpi

Also the download is now to a temporary file followed by an atomic move, which should reduce the chance of corrupted files breaking the installation.

Changelog entry:

• The -name option to the install-plugin command is no longer necessary nor honored.

[daniel-beck]

As a side effect, this will take care of JENKINS-29065 in that "installation" of an invalid file won't work anymore, possibly breaking Jenkins.

[jglick]

"installation" of an invalid file won't work anymore, possibly breaking Jenkins

Confirmed.

[oleg-nenashev]

Looks good to me

[jglick]

Test flake:

java.lang.AssertionError
	at org.junit.Assert.fail(Assert.java:86)
	at org.junit.Assert.assertTrue(Assert.java:41)
	at org.junit.Assert.assertTrue(Assert.java:52)
	at hudson.util.RetrierTest.sleepWorksTest(RetrierTest.java:100)

[oleg-nenashev]

Please add checks for Runtime exceptions when reading the manifest (or just for the entire try/catch block)

addressed

[oleg-nenashev]

It is rather a bug than a feature, but renaming PR retriggers the build more reliably than close/open