Re: svn commit: r1536700 - in /subversion/trunk/subversion/tests: libsvn_client/ libsvn_client/client-test.c libsvn_delta/random-test.c libsvn_diff/ libsvn_diff/diff-diff3-test.c svn_test.h svn_test_main.c

On 10/29/13 6:07 AM, brane_at_apache.org wrote:
> Author: brane
> Date: Tue Oct 29 13:07:49 2013
> New Revision: 1536700
>
> URL: http://svn.apache.org/r1536700
> Log:
> Begin converting C tests to use the same transient directory as the Python
> tests, so that both can be run off a RAM disk.

I gave this a spin (with a r1536831 checkout). Some results of `make check
PARALLEL=1`:

Ubuntu 12.04 on a VM (disk I/O is slow):
subversion/tests on RAM disk: 8:22.25 (m:ss)
subversion/tests/cmdline/svn-test-work on RAM disk: 12:05.16 (m:ss)
subversion/tests/cmdline/svn-test-work on RAM disk (without r1536700): 12:03.78
(m:ss)

OS X 10.7.5 (SSD, disk I/O is fast):
subversion/tests on RAM disk: 9:47.735(m:ss)
subversion/tests/cmdline/svn-test-work on RAM disk: 9:52.231 (m:ss)
subversion/tests/cmdline/svn-test-work on RAM disk (without r1536700): 10:6.924
(m:ss)

So it appears to nominally help. But mounting all of subversion/tests still is
much better for me on my VM with slow I/O. As in the past mounting all of
subversion/tests on OS X isn't worth the added hassle that it entails.

The extra time on the VM run is probably just a fluke. I was only running this
to see if I could ditch the subversion/tests mount on my VM.
Received on 2013-10-29 21:18:28 CET