> On 5 Jan 2019, at 17:03, Alan Bateman <alan.bate...@oracle.com> wrote: > > On 04/01/2019 17:18, Chris Hegarty wrote: >> Will compilations with `--release N-1` wind back the set of allowable >> identifiers? It possibly should, if so how does one do similar when/if >> the set of currency characters is expanded in an update release? >> > I don't think `javac --release` takes the Unicode version into account.
That is a bug. > I agree with your comment that the CSR could be be expanded to at least make > it clear that if support for a currency symbol is added in some future update > of JDK N then source code using it in identifiers will compile with the JDK > update that supports the symbol, not by the GA or previous updates of JDK N. The question is whether, or not, it is reasonable to allow ( part of ) the set of Java identifiers, in the language, to be an implementation detail? If so, then certain code compiled and run on BuzzCorps Java implementation may not compile or run on WollyInc's Java implementation. ( Both are compliant Java SE implementations ) I accept that the set of implementation specific currency characters may noy be all that large, but this does seem like a significant change, and I want to ensure that the implications are fully understood. -Chris.