You are viewing a plain text version of this content. The canonical link for it is here.
Posted to commits@subversion.apache.org by st...@apache.org on 2013/06/14 16:40:18 UTC
svn commit: r1493097 -
/subversion/trunk/subversion/tests/cmdline/svntest/main.py
Author: stsp
Date: Fri Jun 14 14:40:17 2013
New Revision: 1493097
URL: http://svn.apache.org/r1493097
Log:
Run tests with an exclusive lock on working copies. This should reduce test
run time and also ensures that exclusive locking mode is tested.
I've run ra_local and ra_serf tests with this change and got no failures.
In any case, if there were any test failures with exlusive locking mode
enabled, they'd most likely expose bugs in the test suite or Subversion itself.
* subversion/tests/cmdline/svntest/main.py: Enable exclusive locking in the
client configuration used for tests.
Modified:
subversion/trunk/subversion/tests/cmdline/svntest/main.py
Modified: subversion/trunk/subversion/tests/cmdline/svntest/main.py
URL: http://svn.apache.org/viewvc/subversion/trunk/subversion/tests/cmdline/svntest/main.py?rev=1493097&r1=1493096&r2=1493097&view=diff
==============================================================================
--- subversion/trunk/subversion/tests/cmdline/svntest/main.py (original)
+++ subversion/trunk/subversion/tests/cmdline/svntest/main.py Fri Jun 14 14:40:17 2013
@@ -581,6 +581,9 @@ password-stores =
[miscellany]
interactive-conflicts = false
+
+[working-copy]
+exclusive-locking = true
"""
# define default server file contents if none provided
RE: svn commit: r1493097 - /subversion/trunk/subversion/tests/cmdline/svntest/main.py
Posted by Bert Huijben <be...@qqmail.nl>.
> -----Original Message-----
> From: Ivan Zhakov [mailto:ivan@visualsvn.com]
> Sent: woensdag 19 juni 2013 13:50
> To: dev@subversion.apache.org; Stefan Sperling
> Subject: Re: svn commit: r1493097 -
> /subversion/trunk/subversion/tests/cmdline/svntest/main.py
>
> On Fri, Jun 14, 2013 at 6:40 PM, <st...@apache.org> wrote:
> > Author: stsp
> > Date: Fri Jun 14 14:40:17 2013
> > New Revision: 1493097
> >
> > URL: http://svn.apache.org/r1493097
> > Log:
> > Run tests with an exclusive lock on working copies. This should reduce test
> > run time and also ensures that exclusive locking mode is tested.
> >
> > I've run ra_local and ra_serf tests with this change and got no failures.
> > In any case, if there were any test failures with exlusive locking mode
> > enabled, they'd most likely expose bugs in the test suite or Subversion
> itself.
> >
> I don't like this change actually:
> 1. Running tests in different configuration than regular users are
> using bad practice
> 2. It also seems to broke svn benchmarks posted every week, because
> now we get totally different numbers for operations.
>
> Could you please make option to running test suite with exclusive
> locking mode and leave it 'off' by default. Thanks!
+1
Bert
Re: svn commit: r1493097 - /subversion/trunk/subversion/tests/cmdline/svntest/main.py
Posted by Ivan Zhakov <iv...@visualsvn.com>.
On Wed, Jun 19, 2013 at 5:45 PM, Ivan Zhakov <iv...@visualsvn.com> wrote:
> On Wed, Jun 19, 2013 at 5:04 PM, Stefan Sperling <st...@elego.de> wrote:
>> On Wed, Jun 19, 2013 at 03:50:01PM +0400, Ivan Zhakov wrote:
>>> On Fri, Jun 14, 2013 at 6:40 PM, <st...@apache.org> wrote:
>>> > Author: stsp
>>> > Date: Fri Jun 14 14:40:17 2013
>>> > New Revision: 1493097
>>> >
>>> > URL: http://svn.apache.org/r1493097
>>> > Log:
>>> > Run tests with an exclusive lock on working copies. This should reduce test
>>> > run time and also ensures that exclusive locking mode is tested.
>>> >
>>> > I've run ra_local and ra_serf tests with this change and got no failures.
>>> > In any case, if there were any test failures with exlusive locking mode
>>> > enabled, they'd most likely expose bugs in the test suite or Subversion itself.
>>> >
>>> I don't like this change actually:
>>> 1. Running tests in different configuration than regular users are
>>> using bad practice
>>
>> I did think about this before making the change.
>>
>> Your argument can be turned around. If we never test the exclusive
>> locking mode, how can we be sure that it works?
>>
> Just make it optional and someone who interested in this particular
> configuration will use it for testing. Or configure dedicated buildbot
> for that.
>
>> And consider that, if a test passes with exclusive locking, it very
>> likely passes with less restrictive locking. But the reverse is not true!
>> Tests could fail in exclusive locking mode due to bugs in the tests
>> or the code, and we would never see those failures until now.
>>
> No. It's just two different configuration and you cannot say that if
> it pass in one configuration it also doesn't have problems with
> another.
>
>>> 2. It also seems to broke svn benchmarks posted every week, because
>>> now we get totally different numbers for operations.
>>
>> That's unfortunate. But what about things like server-side caching?
>> Don't improvements in such areas have similar effects? I think having
>> better test coverage and test speed is more important than keeping
>> the benchmark results consistent over time.
> Server-side caching is default configuration. I'm just asking your
> keep running test suite in default configuration, which most (at least
> 80%) users are using.
>
>>
>>> Could you please make option to running test suite with exclusive
>>> locking mode and leave it 'off' by default. Thanks!
>>
>> I could do that, yes. But it multiples the number of test configurations
>> yet again, which I don't like. If we do this, I can switch my buildbot
>> to use exclusive locks. And if the buildbot fails some day or we stop
>> maintaining the bot, test coverage will get worse again because nobody
>> tests exclusive mode anymore. So I'd rather have the default be 'on'.
> Multiple tests configurations is great because every person/build can
> right tests for different configuration. But default should be the
> same as default Subversion configuration.
>
>
I've reverted this change in r1496437. You may add test run option if
you'd like to test Subversion with exclusive locking.
--
Ivan Zhakov
CTO | VisualSVN | http://www.visualsvn.com
Re: svn commit: r1493097 - /subversion/trunk/subversion/tests/cmdline/svntest/main.py
Posted by Ivan Zhakov <iv...@visualsvn.com>.
On Wed, Jun 19, 2013 at 5:04 PM, Stefan Sperling <st...@elego.de> wrote:
> On Wed, Jun 19, 2013 at 03:50:01PM +0400, Ivan Zhakov wrote:
>> On Fri, Jun 14, 2013 at 6:40 PM, <st...@apache.org> wrote:
>> > Author: stsp
>> > Date: Fri Jun 14 14:40:17 2013
>> > New Revision: 1493097
>> >
>> > URL: http://svn.apache.org/r1493097
>> > Log:
>> > Run tests with an exclusive lock on working copies. This should reduce test
>> > run time and also ensures that exclusive locking mode is tested.
>> >
>> > I've run ra_local and ra_serf tests with this change and got no failures.
>> > In any case, if there were any test failures with exlusive locking mode
>> > enabled, they'd most likely expose bugs in the test suite or Subversion itself.
>> >
>> I don't like this change actually:
>> 1. Running tests in different configuration than regular users are
>> using bad practice
>
> I did think about this before making the change.
>
> Your argument can be turned around. If we never test the exclusive
> locking mode, how can we be sure that it works?
>
Just make it optional and someone who interested in this particular
configuration will use it for testing. Or configure dedicated buildbot
for that.
> And consider that, if a test passes with exclusive locking, it very
> likely passes with less restrictive locking. But the reverse is not true!
> Tests could fail in exclusive locking mode due to bugs in the tests
> or the code, and we would never see those failures until now.
>
No. It's just two different configuration and you cannot say that if
it pass in one configuration it also doesn't have problems with
another.
>> 2. It also seems to broke svn benchmarks posted every week, because
>> now we get totally different numbers for operations.
>
> That's unfortunate. But what about things like server-side caching?
> Don't improvements in such areas have similar effects? I think having
> better test coverage and test speed is more important than keeping
> the benchmark results consistent over time.
Server-side caching is default configuration. I'm just asking your
keep running test suite in default configuration, which most (at least
80%) users are using.
>
>> Could you please make option to running test suite with exclusive
>> locking mode and leave it 'off' by default. Thanks!
>
> I could do that, yes. But it multiples the number of test configurations
> yet again, which I don't like. If we do this, I can switch my buildbot
> to use exclusive locks. And if the buildbot fails some day or we stop
> maintaining the bot, test coverage will get worse again because nobody
> tests exclusive mode anymore. So I'd rather have the default be 'on'.
Multiple tests configurations is great because every person/build can
right tests for different configuration. But default should be the
same as default Subversion configuration.
--
Ivan Zhakov
CTO | VisualSVN | http://www.visualsvn.com
Re: svn commit: r1493097 -
/subversion/trunk/subversion/tests/cmdline/svntest/main.py
Posted by Stefan Sperling <st...@elego.de>.
On Wed, Jun 19, 2013 at 03:50:01PM +0400, Ivan Zhakov wrote:
> On Fri, Jun 14, 2013 at 6:40 PM, <st...@apache.org> wrote:
> > Author: stsp
> > Date: Fri Jun 14 14:40:17 2013
> > New Revision: 1493097
> >
> > URL: http://svn.apache.org/r1493097
> > Log:
> > Run tests with an exclusive lock on working copies. This should reduce test
> > run time and also ensures that exclusive locking mode is tested.
> >
> > I've run ra_local and ra_serf tests with this change and got no failures.
> > In any case, if there were any test failures with exlusive locking mode
> > enabled, they'd most likely expose bugs in the test suite or Subversion itself.
> >
> I don't like this change actually:
> 1. Running tests in different configuration than regular users are
> using bad practice
I did think about this before making the change.
Your argument can be turned around. If we never test the exclusive
locking mode, how can we be sure that it works?
And consider that, if a test passes with exclusive locking, it very
likely passes with less restrictive locking. But the reverse is not true!
Tests could fail in exclusive locking mode due to bugs in the tests
or the code, and we would never see those failures until now.
> 2. It also seems to broke svn benchmarks posted every week, because
> now we get totally different numbers for operations.
That's unfortunate. But what about things like server-side caching?
Don't improvements in such areas have similar effects? I think having
better test coverage and test speed is more important than keeping
the benchmark results consistent over time.
> Could you please make option to running test suite with exclusive
> locking mode and leave it 'off' by default. Thanks!
I could do that, yes. But it multiples the number of test configurations
yet again, which I don't like. If we do this, I can switch my buildbot
to use exclusive locks. And if the buildbot fails some day or we stop
maintaining the bot, test coverage will get worse again because nobody
tests exclusive mode anymore. So I'd rather have the default be 'on'.
Re: svn commit: r1493097 - /subversion/trunk/subversion/tests/cmdline/svntest/main.py
Posted by Ivan Zhakov <iv...@visualsvn.com>.
On Fri, Jun 14, 2013 at 6:40 PM, <st...@apache.org> wrote:
> Author: stsp
> Date: Fri Jun 14 14:40:17 2013
> New Revision: 1493097
>
> URL: http://svn.apache.org/r1493097
> Log:
> Run tests with an exclusive lock on working copies. This should reduce test
> run time and also ensures that exclusive locking mode is tested.
>
> I've run ra_local and ra_serf tests with this change and got no failures.
> In any case, if there were any test failures with exlusive locking mode
> enabled, they'd most likely expose bugs in the test suite or Subversion itself.
>
I don't like this change actually:
1. Running tests in different configuration than regular users are
using bad practice
2. It also seems to broke svn benchmarks posted every week, because
now we get totally different numbers for operations.
Could you please make option to running test suite with exclusive
locking mode and leave it 'off' by default. Thanks!
--
Ivan Zhakov
CTO | VisualSVN | http://www.visualsvn.com