RFR: 8196750 [Testbug] tools/launcher tests need to tolerate unrelated warnings

[David Holmes]

On 22/03/2018 12:52 PM, Andrei Nazarov wrote:
> On Mar 21, 2018 19:13, David Holmes <david.holmes at oracle.com> wrote:
> 
>     Hi Andrei,
> 
>     On 22/03/2018 11:12 AM, Andrey Nazarov wrote:
>      > Hi,
>      >
>      > Please review fix in launcher tests.
>      >
>      > Review: http://cr.openjdk.java.net/~anazarov/JDK-8196750/webrev/
>      > Bug: https://bugs.openjdk.java.net/browse/JDK-8196750?filter=-1
> 
>     test/jdk/tools/launcher/ToolsOpts.java
> 
>     I don't understand how the change fixes the problem. IIUC the
>     problem is
>     that the test expects the output to consist of the command-line passed
>     to the tool. Instead if it contains a Warning from the VM, it won't
>     match and so we fail. The new code no longer uses a String[] and no
>     longer assumes that output[i] == args[i],
> 
> output now is List<String> and
> output.get(j).equals(arg[j]) has the same sematics.
> The difference is that it *searches* args as a line in the output, so it 
> would skip any Warnings lines.

So for each arg it searches the entire output.

Okay I see now.

Thanks,
David
-----
> Output is scanned once so it checks order of args as well.
> 
> -- Andrei
> 
>     it still searches the
>     current "output" value for a match with one of the args. But that will
>     still fail if what we actually have is a Warning. ?? I would have
>     expected to see the Warning strings filtered out more directly.
> 
>     The other test change seems reasonable.
> 
>     Thanks,
>     David
> 
>      > —Thanks,
>      >   Andrei
>      >
> 
>