RFR(XS): 8160457: VersionProps.versionNumbers() is broken

[Mandy Chung]

> On Jun 28, 2016, at 7:57 AM, Volker Simonis <volker.simonis at gmail.com> wrote:
> 
> Hi,
> 
> can somebody please review this trivial fix:
> 
> http://cr.openjdk.java.net/~simonis/webrevs/2016/8160457/
> https://bugs.openjdk.java.net/browse/JDK-8160457
> 

+1

> The problem is that VersionProps.versionNumbers() incorrectly parses a
> Java version string because it doesn't skip the separating dots. The
> current version only works for one digit versions like "9" but will
> fail for any longer version string like for example "9.0.0.1" with:
> 
>        at java.lang.NumberFormatException.forInputString(java.base at 9.0.0.1-internal/NumberFormatException.java:65)
>        at java.lang.Integer.parseInt(java.base at 9.0.0.1-internal/Integer.java:791)
>        at java.lang.VersionProps.versionNumbers(java.base at 9.0.0.1-internal/VersionProps.java:76)
>        at java.lang.Runtime.version(java.base at 9.0.0.1-internal/Runtime.java:940)
> 
> This also breaks the build which uses a newly built jdk for
> bootstrapping if we set '--with-version-patch=1' for example.
> 
> An can you PLEASE, PLEASE finally do your internal/early access builds
> with '--with-version-patch=1’.

Iris and I both have asked to verify the builds with --with-version-patch.  

Claes - did you see the exception during your verification?

> It seems really careless to me that you
> introduce a new, up to four digit versioning schema but only test the
> shortcut version with one digit.

I brought up similar question on the testing side.  Automatic builds configured to build different version number could catch this and should be considered to do. 

Mandy

> I wouldn't be surprised if a version
> like "9.0.0.1" breaks more Java applications than the sun.misc.Unsafe
> removal :)
> 
> Thank you and best regards,
> Volker