Upgrade commons libs, asm, jgit, jna, jackson, felix, jacoco and junit

[mbien]

tried to find some low hanging fruits using the dependency checker: https://github.com/apache/netbeans/actions/runs/8543693772

one commit per library, if I should drop a commit please say so.

• Upgrade jsch from 0.1.72 to 0.2.17 outside of jgit*s declared range
• Upgrade jackson from 2.16.1 to 2.17.0
• Upgrade jacoco.core from 0.8.7 to 0.8.12
• Upgrade junit from 5.6.0 to 5.10.2
• Upgrade JNA from 5.12.1 to 5.14.0
• Upgrade felix from 7.0.3 to 7.0.5
• Upgrade jgit from 9.8.0 to 9.9.0
  • Upgrade JavaEWAH from 1.1.6 to 1.2.3
• Upgrade ASM from 9.6 to 9.7
• Upgrade commons-compress from 1.25.0 to 1.26.1
• Upgrade commons-logging from 1.3.0 to 1.3.1
• Upgrade commons-io from 2.15.1 to 2.16.0
• Upgrade commons-codec from 1.16.0 to 1.16.1

[ebarboni]

Nice.
What is the issue with jsch ?

[mbien]

Nice. What is the issue with jsch ?

I should have checked what version jgit needs first before updating that. Jgit is loaded as osgi module and it won't load dependencies which are outside the declared range -> classloader error. The manifest has com.jcraft.jsch;version="[0.1.37,0.2.0)" in it, so I updated out of range by accident.

mvnd eu.maveniverse.maven.plugins:toolbox:gav-tree -Dgav=org.eclipse.jgit:org.eclipse.jgit.ssh.jsch:6.9.0.202403050737-r
[INFO] --------------------------------[ pom ]---------------------------------
[INFO] 
[INFO] --- toolbox:0.1.3:gav-tree (default-cli) @ roller-project ---
[INFO] org.eclipse.jgit:org.eclipse.jgit.ssh.jsch:jar:6.9.0.202403050737-r
[INFO] +- org.eclipse.jgit:org.eclipse.jgit:jar:6.9.0.202403050737-r [compile]
[INFO] |  +- com.googlecode.javaewah:JavaEWAH:jar:1.2.3 [compile]
[INFO] |  \- commons-codec:commons-codec:jar:1.16.0 [compile]
[INFO] +- com.jcraft:jsch:jar:0.1.55 [compile]
[INFO] +- com.jcraft:jzlib:jar:1.1.3 [compile]
[INFO] \- org.slf4j:slf4j-api:jar:1.7.36 [compile]
[INFO] ------------------------------------------------------------------------

speaking of which... I should probably update JavaEWAH too.

edit: have to investigate that failing php test on windows

[mbien]

PHP test failure on win is likely file path length / max cp length / max arg size related. The forked JVM can't start. Will try to debug this on windows and check what is causing this.

alternative is to drop the commons-io commit since this seems to be sufficient to "resolve" this.

edit: suspicion confirmed, it is not the file path length itself, it is the gigantic classpath combined with the path prefix which goes over the process arg length limit. I could still make it test if I moved the workspace closer to the drive letter, going to try to replicate it in gh actions.

[matthiasblaesing]

Looks sane to me. Only eyeballed this.

[mbien]

the interesting aspect about this issue here was that the test support classes take the first completion item and simply go with it (in this case 'methodTag' and 'methodTest'). It seems like what is first can depend on the environment which can cause test failures. For some reason the lib upgrade here guaranteed test failure on windows in this particular php test case.

netbeans/ide/csl.api/test/unit/src/org/netbeans/modules/csl/api/test/CslTestBase.java

Lines 3148 to 3154 in cac7b2c

	CompletionProposal match = null;
	for (CompletionProposal proposal : proposals) {
	if (proposal.getName().startsWith(itemPrefix)) {
	match = proposal;
	break;
	}
	}

I think we should consider to fail the test right away when this loop returns more than one match since this would indicate that the test case is likely not stable and then update the tests. Or make sure that proposals has a stable order.

[mbien]

did some manual testing and everything worked fine, all tests green -> merging