Re: [xml] XML regression test cases...

["Jain, Nilesh" <nilesh jain intel com>]

On Thu, 2005-10-20 at 17:44 -0400, Daniel Veillard wrote: 
> On Thu, Oct 20, 2005 at 02:27:54PM -0700, Jain, Nilesh wrote:
> > I completely understand the point you have mentioned, but same time my
> > real concern is for application/users of libxml2... because if
> > application is depends or based on certain behavior of library, which
> > changes release over release [I understand potential reasons for change]
> > then how application deals with that.. do they change the application...
> > or applications are usually transparent from these changes. 
>
>   applications are usually transparent from these changes

That solve my lot of worries 

> > I guess my real question here is to understand the impact of behavior
> > changes release over release on applications, and how they deal with
> > that?
>
>    make correct user code, i.e. catch errors, usually they fon't have
> to deal specifically with it, a tree is way more complex than say an int,
> you already have to make flexible code to deal with trees anyway.
>
> > Or is it certain section/module of library which can vary
> > depending on the underling specification interpretation change, which is
> > really transparent to application... 
>
>   that too, depends on the kind of fixes
>
> > I am here not questioning the changes but same time trying to understand
> > the impact on application, which LSB cares most... I am putting this
> > question here as you have more visibility of usage and impact of these
> > changes...  
>
>   And I am answering that when I do such change I carefully think about 
> how applications would react, I'm sure I said that at least a couple of time
> already. I can't dump my knowledge in your brain, there is collective knowledge
> on this list about libxml2 and their apps, I think they somehow at least trust
> me from not breaking their apps, and I use those various tests suites
> accumulated over the years to raise my level of trust on my own changes (or the
> changes from other contributors).
>   You can do a very dumb test like upgrading libxml2 and libxslt on a RHEL3
> system to the latest RPMs I ship and chase for breakages in applications,
> if you have specific application compatibilities, test them, or at least
> list them.

Here I am neither mistrusting you nor challenging your creditability. I
have a complete trust on you, but trying to understand the difference I
see. My concern started when I ran the older version of internal
regression test against the newer version of library to ensure behavior
stability, and there I see some variation. I was assuming that these
test are only testing APIs visible to application, my assumption was
wrong. So in this case my conclusion is that those test cases are also
testing part of functionality which is not directly visible to
application. Is my understanding correct.

Thanks,
Nilesh