https://issues.apache.org/bugzilla/show_bug.cgi?id=54128

--- Comment #25 from Tim Whittington <[email protected]> ---
> It seems implausible for the entire test to execute in 1 second, bettering
> the execution time of the 'type' command.

Having said that, I can get the execution time down to about 2 seconds with Ant
1.7.1 by piping the output to a file or the NUL device.

With that approach Ant 1.7.1 is clearly faster (about 2 seconds) than 1.9.3+
(about 4 seconds) which I imagine will be the 100ms pause penalty applied in
StreamPumper.waitForInput(InputStream) each time there is 0 bytes available
(which is often the case after reading a line from STDIN).

Some experimentation reveals that reducing that pause time does speed up the
1.9.x execution, but if it's put too low the execution times ramp up a lot - it
appears that there's a significant penalty to calling available() too much. The
penalty isn't in calling available() (that takes about 10-20 microseconds on
average), so it might some kind of contention.

-- 
You are receiving this mail because:
You are the assignee for the bug.

Reply via email to