Re: svn commit: r31084 - branches/1.5.x

[Karl Fogel <kf...@red-bean.com>]

danielsh@tigris.org writes:
> Log:
> Nominate r31083.

Not a big deal, but is it really necessary to port new tests over to a
release/maintenance branch?  When they test something that was
specifically ported to that branch, I can understand it, but except for
that case, my feeling is let's not make more work for the release
manager (and reviewers).

I'm certainly prepared to be shown that I haven't thought this through
carefully enough, of course... :-)

-Karl

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org

Re: svn commit: r31084 - branches/1.5.x

[Daniel Shahaf <da...@tigris.org>]

Karl Fogel wrote on Thu, 8 May 2008 at 15:22 -0400:
> Daniel Shahaf <da...@tigris.org> writes:
> > following discussion I moved the nomination to 1.5.1.  Does this address 
> > your concerns?
> 
> Well, my concerns were more along the lines of "Is this worth porting to
> 1.5.x at all?"  Since svn_path_is_canonical() is new in 1.5, there's a
> better argument for porting this test than most tests; however, it's
> still going to get tested on trunk (1.6, 1.7, etc) all the time anyway,
> so I personally don't think it's worth the porting overhead.
> 
> But it's a judgement call.  I'm not asking you to remove it from STATUs,
> just reminding that porting comes at a cost (in reviewing/voting, and in
> merging by whoever does the merge), so let's not go overboard.
> 

Since currently the implementations on trunk and in 1.5.x are identical,
and since 1.5.x has tests for svn_path_canonicalize(), I'll unnominate
it in r31099.

Thanks for the reminder,

Daniel

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org

Re: svn commit: r31084 - branches/1.5.x

[Karl Fogel <kf...@red-bean.com>]

Daniel Shahaf <da...@tigris.org> writes:
> In this case, I agree that the test is not strictly necessary for 1.5.x,
> since svn_path_is_canonical is a trivial wrapper around
> svn_path_canonicalize, and the latter is already tested.
>
>     svn_boolean_t
>     svn_path_is_canonical(const char *path, apr_pool_t *pool)
>     {
>       return (strcmp(path, svn_path_canonicalize(path, pool)) == 0);
>     }
>
> In general, however, I assumed that tests for an API that made it to
> a release branch should be backported to the branch (whether or not the
> API is new in that branch).
>
> Stefan Sperling also questioned the backport over IRC this afternoon; 
> following discussion I moved the nomination to 1.5.1.  Does this address 
> your concerns?

Well, my concerns were more along the lines of "Is this worth porting to
1.5.x at all?"  Since svn_path_is_canonical() is new in 1.5, there's a
better argument for porting this test than most tests; however, it's
still going to get tested on trunk (1.6, 1.7, etc) all the time anyway,
so I personally don't think it's worth the porting overhead.

But it's a judgement call.  I'm not asking you to remove it from STATUs,
just reminding that porting comes at a cost (in reviewing/voting, and in
merging by whoever does the merge), so let's not go overboard.

-K

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org

Re: svn commit: r31084 - branches/1.5.x

[Daniel Shahaf <da...@tigris.org>]

Karl Fogel wrote on Thu, 8 May 2008 at 14:50 -0400:
> danielsh@tigris.org writes:
> > Log:
> > Nominate r31083.
> 
> Not a big deal, but is it really necessary to port new tests over to a
> release/maintenance branch?  When they test something that was
> specifically ported to that branch, I can understand it, but except for
> that case, my feeling is let's not make more work for the release
> manager (and reviewers).
> 
> I'm certainly prepared to be shown that I haven't thought this through
> carefully enough, of course... :-)
> 
> -Karl
> 

In this case, I agree that the test is not strictly necessary for 1.5.x,
since svn_path_is_canonical is a trivial wrapper around
svn_path_canonicalize, and the latter is already tested.

    svn_boolean_t
    svn_path_is_canonical(const char *path, apr_pool_t *pool)
    {
      return (strcmp(path, svn_path_canonicalize(path, pool)) == 0);
    }

In general, however, I assumed that tests for an API that made it to
a release branch should be backported to the branch (whether or not the
API is new in that branch).

Stefan Sperling also questioned the backport over IRC this afternoon; 
following discussion I moved the nomination to 1.5.1.  Does this address 
your concerns?


Thanks,

Daniel

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org