You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@subversion.apache.org by Ivan Zhakov <ch...@gmail.com> on 2006/04/12 06:24:31 UTC
Re: svn commit: r19317 - trunk/subversion/tests/cmdline/svntest
On 4/12/06, mhagger@tigris.org <mh...@tigris.org> wrote:
> Author: mhagger
> Date: Tue Apr 11 22:31:41 2006
> New Revision: 19317
>
> Modified:
> trunk/subversion/tests/cmdline/svntest/main.py
>
> Log:
> If a test skips, clean up its sandbox.
>
> Previously only tests whose converted results indicated a pass had
> their sandboxes cleaned up.
>
> * subversion/tests/cmdline/svntest/main.py (TestRunner.run): Only omit
> sandbox cleanup for tests that result in a failure (not tests that
> skip).
>
>
I'm not sure that it's right change. It is frequent use-case retry
last operation by hand or look to working copy, when test failed.
--
Ivan Zhakov
Re: svn commit: r19317 - trunk/subversion/tests/cmdline/svntest
Posted by Ivan Zhakov <ch...@gmail.com>.
On 4/13/06, Daniel Rall <dl...@collab.net> wrote:
> On Wed, 12 Apr 2006, Julian Foad wrote:
>
> > Michael Haggerty wrote:
> > >But if a test causes itself to be skipped (by raising an svntest.Skip
> > >exception) then there wouldn't be much call for running it again, would
> > >there? The sandboxes of tests that fail are *not* cleaned up, even
> > >after this commit. (My guess was that the old behavior was an oversight
> > >in the implementation of the skip behavior.)
> >
> > Agreed.
> >
> > >But I would certainly be happy to restore the old behavior if that is
> > >what people want.
> >
> > No.
> >
> > >On a related point, if a test is explicitly marked Skip (by being listed
> > >as Skip(test_name, 1)) then it seems superfluous to generate a sandbox
> > >in the first place. (For tests that are skipped by throwing a Skip
> > >exception, on the other hand, it is not so easy to avoid creating a
> > >Sandbox.) So I was thinking of skipping sandbox creation for explicitly
> > >Skipped tests. What do people think of that?
> >
> > +0: not a significant time saving, as all "Sandbox::__init__" does is set a
> > few variables. The significant time taken to build a sandbox is in the
> > "sandbox.build()" call within a test.
>
> I concur with Julian's comments.
>
> Avoiding creation of a Sandbox object would be nice if it can be
> handled with a minimal amount of additional code.
Agreed with you and Julian. I didn't investigate in commit content.
--
Ivan Zhakov
Re: svn commit: r19317 - trunk/subversion/tests/cmdline/svntest
Posted by Michael Haggerty <mh...@alum.mit.edu>.
Daniel Rall wrote:
> On Wed, 12 Apr 2006, Julian Foad wrote:
>>Michael Haggerty wrote:
>>>On a related point, if a test is explicitly marked Skip (by being listed
>>>as Skip(test_name, 1)) then it seems superfluous to generate a sandbox
>>>in the first place. (For tests that are skipped by throwing a Skip
>>>exception, on the other hand, it is not so easy to avoid creating a
>>>Sandbox.) So I was thinking of skipping sandbox creation for explicitly
>>>Skipped tests. What do people think of that?
>
> Avoiding creation of a Sandbox object would be nice if it can be
> handled with a minimal amount of additional code.
Yes, it is a clean change requiring only one "if", so I committed this
change as r19401.
Michael
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: svn commit: r19317 - trunk/subversion/tests/cmdline/svntest
Posted by Daniel Rall <dl...@collab.net>.
On Wed, 12 Apr 2006, Julian Foad wrote:
> Michael Haggerty wrote:
> >But if a test causes itself to be skipped (by raising an svntest.Skip
> >exception) then there wouldn't be much call for running it again, would
> >there? The sandboxes of tests that fail are *not* cleaned up, even
> >after this commit. (My guess was that the old behavior was an oversight
> >in the implementation of the skip behavior.)
>
> Agreed.
>
> >But I would certainly be happy to restore the old behavior if that is
> >what people want.
>
> No.
>
> >On a related point, if a test is explicitly marked Skip (by being listed
> >as Skip(test_name, 1)) then it seems superfluous to generate a sandbox
> >in the first place. (For tests that are skipped by throwing a Skip
> >exception, on the other hand, it is not so easy to avoid creating a
> >Sandbox.) So I was thinking of skipping sandbox creation for explicitly
> >Skipped tests. What do people think of that?
>
> +0: not a significant time saving, as all "Sandbox::__init__" does is set a
> few variables. The significant time taken to build a sandbox is in the
> "sandbox.build()" call within a test.
I concur with Julian's comments.
Avoiding creation of a Sandbox object would be nice if it can be
handled with a minimal amount of additional code.
--
Daniel Rall
Re: svn commit: r19317 - trunk/subversion/tests/cmdline/svntest
Posted by Julian Foad <ju...@btopenworld.com>.
Michael Haggerty wrote:
> But if a test causes itself to be skipped (by raising an svntest.Skip
> exception) then there wouldn't be much call for running it again, would
> there? The sandboxes of tests that fail are *not* cleaned up, even
> after this commit. (My guess was that the old behavior was an oversight
> in the implementation of the skip behavior.)
Agreed.
> But I would certainly be happy to restore the old behavior if that is
> what people want.
No.
> On a related point, if a test is explicitly marked Skip (by being listed
> as Skip(test_name, 1)) then it seems superfluous to generate a sandbox
> in the first place. (For tests that are skipped by throwing a Skip
> exception, on the other hand, it is not so easy to avoid creating a
> Sandbox.) So I was thinking of skipping sandbox creation for explicitly
> Skipped tests. What do people think of that?
+0: not a significant time saving, as all "Sandbox::__init__" does is set a few
variables. The significant time taken to build a sandbox is in the
"sandbox.build()" call within a test.
- Julian
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: svn commit: r19317 - trunk/subversion/tests/cmdline/svntest
Posted by Michael Haggerty <mh...@alum.mit.edu>.
Ivan Zhakov wrote:
> On 4/12/06, mhagger@tigris.org <mh...@tigris.org> wrote:
>
>>Author: mhagger
>>Date: Tue Apr 11 22:31:41 2006
>>New Revision: 19317
>>
>>Modified:
>> trunk/subversion/tests/cmdline/svntest/main.py
>>
>>Log:
>>If a test skips, clean up its sandbox.
>>
>>Previously only tests whose converted results indicated a pass had
>>their sandboxes cleaned up.
>>
>>* subversion/tests/cmdline/svntest/main.py (TestRunner.run): Only omit
>> sandbox cleanup for tests that result in a failure (not tests that
>> skip).
>
> I'm not sure that it's right change. It is frequent use-case retry
> last operation by hand or look to working copy, when test failed.
But if a test causes itself to be skipped (by raising an svntest.Skip
exception) then there wouldn't be much call for running it again, would
there? The sandboxes of tests that fail are *not* cleaned up, even
after this commit. (My guess was that the old behavior was an oversight
in the implementation of the skip behavior.)
But I would certainly be happy to restore the old behavior if that is
what people want.
On a related point, if a test is explicitly marked Skip (by being listed
as Skip(test_name, 1)) then it seems superfluous to generate a sandbox
in the first place. (For tests that are skipped by throwing a Skip
exception, on the other hand, it is not so easy to avoid creating a
Sandbox.) So I was thinking of skipping sandbox creation for explicitly
Skipped tests. What do people think of that?
Yours,
Michael
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org