On 27/04/2018 19:23, Rafael Winterhalter wrote:
Hei Alan and David,
thanks for pointing me to the issue, I have only searched the release
notes for u172 by accident.
The issue is mainly during builds. I run my library on multiple CI
servers to cover Windows/Linux and different Java versions from 6-11.
Unfortunately, I have not full control over what version of Java is
run on these servers. Yesterday, I found some of the builds fail for
pull requests what was a bit confusing. Byte Buddy offers an
abstraction over type descriptions that implement similar semantics to
the Java reflection API when it comes to equality and to textual
(toString) representations. These tests suddenly failed since the JVMs
representation is changed, this is all. The Scala build had a similar
problem: https://github.com/scala/bug/issues/10835
This is not a big deal but I found it surprising to have a change in
the string representation within an update release. Especially since a
nested class does not necessarily have the same name prefix if a class
is not compiled with javac. I would have preferred the consistency
over the redundancy; especially when type names are machine-processed,
this often makes a parser easier to implement.
There isn't sufficient information in the bug to understand why it was
back-ported to 8u172. That said, aren't you replying on unspecified
behavior? I don't think toString is specified here.
-Alan
