RFR: JDK-8232834: RunTest sometimes fails to produce valid exitcode.txt

[Magnus Ihse Bursie]

On 2019-10-23 14:41, Erik Joelsson wrote:
>
> Yes, I've been torn about this. On Windows I made measurable build 
> speed improvements some years ago by minimizing the extra shell 
> invocations for compile commands, so I didn't like the idea of 
> introducing an unconditional extra shell. Another option could be to 
> scan the input for '[<>]' in ExecuteWithLog and only add it if found. 
> It will certainly make the ExecuteWithLog body convoluted but may be a 
> better solution.
>
Yes, that might be best. If we write a support function 
AddSubshellIfNeeded or so, and wrap the command line in it, it might be 
okay-ish. :)

I filed https://bugs.openjdk.java.net/browse/JDK-8233115 to keep track 
of this.

/Magnus