[REPL] Stop old classpath calculation on the base classloader #4363
Conversation
ileasile
force-pushed
the
rr/ileasile/border-guard-classloader
branch
from
f1a7916 to
6893731
Compare
ileasile
force-pushed
the
rr/ileasile/border-guard-classloader
branch
from
6893731 to
a07c7a7
Compare
ileasile
changed the title
ligee
left a comment
ligee
left a comment
There was a problem hiding this comment.
Please add some info in the commit message, e.g. "since the mechanism of the recursive classpath checking is intended for the "inner" scripting/REPL classloaders, and should not touch anything beyond."
| fun recursiveClassPath(res: Sequence<URL>, classLoader: ClassLoader?): Sequence<URL> = | ||
| when (classLoader) { | ||
| null -> res | ||
| null, baseClassLoader -> res |
There was a problem hiding this comment.
I would still process the null case, e.g. generating an error if we've got null.
There was a problem hiding this comment.
But it should be after baseClassLoader check, since it could be null, and it is a valid situation.
There was a problem hiding this comment.
@ligee I'm not sure that fallbackClassloader of DualClassLoader should necessarily have base classloader as a parent. Is it really so?
There was a problem hiding this comment.
As I can see from usages, it is always a thread context classloader that is a base classloader at the same time, so there is no appropriate test
There was a problem hiding this comment.
I'd say it should convert to the baseClassLoader, but since it is possible to replace it, it is safer to leave the code as you wrote it.
About the tests - there are some tests with baseClassLoader set to null, and that's probably it.
|
Merged manually: bab5d16 |
Motivation for this change is given in the comment.