Fix "URI is not hierarchical" during runtime attachment

[jeanbisutti]

When the runtime attachment feature was used with a Spring Boot application (jar), the runtime attachement dit not work and the following stack trace appeared:

    Caused by: java.lang.IllegalArgumentException: URI is not hierarchical
    	at java.base/java.io.File.<init>(File.java:420)
    	at io.opentelemetry.contrib.attach.AgentFileLocator.locateAgentFile(AgentFileLocator.java:25)
    	at io.opentelemetry.contrib.attach.RuntimeAttach.attachJavaagentToCurrentJVM(RuntimeAttach.java:18)

This PR fixes the issue.

[trask]

is there a way to know whether copying is needed, and only copy in that case?

[trask]

I would think that the JVM would still need access to the jar file for lazy loading additional classes? (in which case we could delete on exit instead)

[jeanbisutti]

Not sure because the Javadoc of Byte Buddy's attach(File agentJar, String processId) method says "This operation blocks until the attachment is complete". In case of doubt, I will delete in exit.

[jeanbisutti]

is there a way to know whether copying is needed, and only copy in that case?

The problem has appeared when the application was packaged as a jar file (I expect the same problem with a war file). The only solution I have found is to use a temp jar file (I have tried to write code with real path but without success). The same thing is done in Elastic agent. We could use a try/catchwith new File(URI) to only copy when it is needed. But I see several drawbacks to this approach:

• It would only be related to the case where the code is executed in an IDE
• It would add more code to maintain
• The fast tests of this PR would not use the same execution path as in production

[trask]

makes sense, we may want to revisit later due to users on read-only file systems, but I agree with keeping it simple for now

[jeanbisutti]

makes sense, we may want to revisit later due to users on read-only file systems, but I agree with keeping it simple for now

We may use the install() method of Byte Buddy

[trask]

can these be null? if so, let's add @Nullable to the params

[jeanbisutti]

This can't be null but i was afraid by build checks 😅

[trask]

@jeanbisutti looks like this needs a spotlessApply then ready for merge!

I pushed spotlessApply 👍

[jeanbisutti]

I pushed spotlessApply 👍

Thank you! Maybe it would be possible to automatically add a spotless:apply commit in the CI