Java launcher with `--module=name[/entry-point]` fails on Windows

[Daniel D. Daugherty]

Forwarding over to core-libs-dev at ...  I think that's where launcher
issues are addressed.

Bcc'ing jdk-dev at ...

Dan


On 11/15/19 6:23 AM, Christian Stein wrote:
> Hi,
>
> relying on the "Note: To specify an argument for a long option,
> you can use either --name=value or --name value." from [0] and
> assuming that `--module` is a long option (with `-m` being its
> short counterpart) I think I found an error in `java.exe`.
>
> The brief history of the search is recorded at JBS [1], and
> yesterday, I managed to create a minimal, complete and
> verifiable example at [2] -- which contains a README.md
> file with more details. Here's a gist:
>
> Calling `java --module-path=lib --module=m/Main --help ...`
> sometimes fails to launch the JVM. The failure, if it occurs,
> expresses mostly via a silent exit of the `java.exe` program
> leaving a negative value in the "DOS shell" `ERRORLEVEL`
> environment variable. Seldom, a fatal error log is written
> to the current working directory.
>
> The underlying reason seems to be an error in the
> calculation of the application argument count and index.
> The later reads: AppArgIndex: -1 points to (null)
>
> Work-around: omit the `=` (equals sign) after `--module`
> and the entry-point, here `m/Main`. This command line
> always gets the JVM up and running.
>
> Cheers,
> Christain
>
> [0] https://docs.oracle.com/en/java/javase/11/tools/java.html
> [1] https://bugs.openjdk.java.net/browse/JDK-8234076
> [2] https://github.com/sormuras/JDK-8234076
>