You are viewing a plain text version of this content. The canonical link for it is here.
Posted to rivet-dev@tcl.apache.org by Valery Masiutsin <va...@gmail.com> on 2007/10/19 18:22:07 UTC

tests in test directory

Hello,  guys !

During the next week i'll try to concentrate my attention on tcl/rvt
test scripts
that could be found in tests directory.
First of all it needs to be tuned for apache 2 support, after that
i'll need to  revue the scripts, probably  add a number of new ones.
The purpose of this is to create some sort of automated suite which
will allow us to test different aspects of rivet2 module. And will
clear show up what does not work, and will make easy to catch up
regressions.
All thoughts on this topic are welcome.

Regards Valery

---------------------------------------------------------------------
To unsubscribe, e-mail: rivet-dev-unsubscribe@tcl.apache.org
For additional commands, e-mail: rivet-dev-help@tcl.apache.org


Re: tests in test directory

Posted by Massimo Manghi <ma...@unipr.it>.
Hi Valery,

As usual your opinions exhibit strong points.

I accept your point of view if Rivet must have new and
absolutely necessary features that are missing now and for which
we don't have any other option except for bringing them into
a future architecture. I mean feature for the application level
For this development we may add new  features for the apache1.x
module developing in pure Tcl or use  the strategy of encapsulating
them into packages (like in the package Rivet that maybe should be
renamed as Rivetlib) At that level adding more feature to the 1_0
should be easy.

Valery Masiutsin wrote:
> Hello, Massimo !
>   
> I disagree. I think we should support  apache-1.3.x version, as long
> as it will be used.
> And it will be used, because apache-1.3.x is the only version of
> apache that has been properly audited for security holes, and spent
> enough time in world wide production to be considered sort of
> reliable.
>  I have nothing against neither  apache2, nor by against folks who are
> using or developing it  but here is my short list of contras:
> -- first and most important - apache2 is a bad known by its api
> changes which break backward compatibility with in  apache2 itself,
> just recall apache 2.0 - apache 2.2 switch.
>   
this is sad, but we have to put up with it.

> -- second introduction of this threading stuff - very questionable
> from my point of view, of course there are some benefits on platforms
> with expensive process creation, but on unix - it just makes things
> much more complicated, threading support varies between different
> common unixes greatly.
>   
this looks like the focal point of every future development. Threads were
on the heap years ago and now everybody seem to have fallen out with them.
Even though I still like the concept I won't make any advocacy, lest a war
of religion would break out ;-)

> What i am thinking about is to make rivet so called "apr-centric" not
> "apache-centric" - most of action code should be redone using apr and
> tcl functions and made apache independent, leaving apache dependent
> bits as thin as it will be possible.  IMHO 2.0-2.2 apache API breakage
> was not the last one :)  I dont want to be dependant on it in the
> future.
>   
seriously this can happen again? Maybe Rivet made the right thing:
pausing the development and waiting for Apache3.0 :-)

 ciao

 -- Massimo


---------------------------------------------------------------------
To unsubscribe, e-mail: rivet-dev-unsubscribe@tcl.apache.org
For additional commands, e-mail: rivet-dev-help@tcl.apache.org


Re: tests in test directory

Posted by David Welton <da...@gmail.com>.
> What i am thinking about is to make rivet so called "apr-centric" not
> "apache-centric" - most of action code should be redone using apr and
> tcl functions and made apache independent, leaving apache dependent
> bits as thin as it will be possible.  IMHO 2.0-2.2 apache API breakage
> was not the last one :)  I dont want to be dependant on it in the
> future.

That's actually how it is done, mostly.  I attempted to hive off as
much as possible into the TclWebapache.c file so that that and
mod_rivet.c are the places where you'd have to change things.

-- 
David N. Welton
http://www.welton.it/davidw/

---------------------------------------------------------------------
To unsubscribe, e-mail: rivet-dev-unsubscribe@tcl.apache.org
For additional commands, e-mail: rivet-dev-help@tcl.apache.org


Re: tests in test directory

Posted by Valery Masiutsin <va...@gmail.com>.
Hello, Massimo !
> Having just apache-2 to take care of would give much more
> freedom to the development and simplify the tests.
>
> David, you're the policy maker here, what do you think?
> Maybe 1_0 is stable enough to be the final release for
> apache1.x?
>
>  -- Massimo
I disagree. I think we should support  apache-1.3.x version, as long
as it will be used.
And it will be used, because apache-1.3.x is the only version of
apache that has been properly audited for security holes, and spent
enough time in world wide production to be considered sort of
reliable.
 I have nothing against neither  apache2, nor by against folks who are
using or developing it  but here is my short list of contras:
-- first and most important - apache2 is a bad known by its api
changes which break backward compatibility with in  apache2 itself,
just recall apache 2.0 - apache 2.2 switch.
-- second introduction of this threading stuff - very questionable
from my point of view, of course there are some benefits on platforms
with expensive process creation, but on unix - it just makes things
much more complicated, threading support varies between different
common unixes greatly.
What i am thinking about is to make rivet so called "apr-centric" not
"apache-centric" - most of action code should be redone using apr and
tcl functions and made apache independent, leaving apache dependent
bits as thin as it will be possible.  IMHO 2.0-2.2 apache API breakage
was not the last one :)  I dont want to be dependant on it in the
future.


Regards Valery.

---------------------------------------------------------------------
To unsubscribe, e-mail: rivet-dev-unsubscribe@tcl.apache.org
For additional commands, e-mail: rivet-dev-help@tcl.apache.org


Re: tests in test directory

Posted by Massimo Manghi <ma...@unipr.it>.
It also may help to keep the development of the apache-1 and
apache-2 together, at least to fix their behavior in
well defined common tests as long as we imagine the 2
modules as providing the same functionalities in 2
different webservers. After all is this what we want?
Having just apache-2 to take care of would give much more 
freedom to the development and simplify the tests.

David, you're the policy maker here, what do you think?
Maybe 1_0 is stable enough to be the final release for
apache1.x?

 -- Massimo


On Fri, 19 Oct 2007 19:22:07 +0300, Valery Masiutsin wrote
> Hello,  guys !
> 
> During the next week i'll try to concentrate my attention on tcl/rvt
> test scripts
> that could be found in tests directory.
> First of all it needs to be tuned for apache 2 support, after that
> i'll need to  revue the scripts, probably  add a number of new ones.
> The purpose of this is to create some sort of automated suite which
> will allow us to test different aspects of rivet2 module. And will
> clear show up what does not work, and will make easy to catch up
> regressions.
> All thoughts on this topic are welcome.
> 
> Regards Valery
> 


--
Universita' degli Studi di Parma (http://www.unipr.it)


---------------------------------------------------------------------
To unsubscribe, e-mail: rivet-dev-unsubscribe@tcl.apache.org
For additional commands, e-mail: rivet-dev-help@tcl.apache.org


Re: tests in test directory

Posted by David Welton <da...@gmail.com>.
> During the next week i'll try to concentrate my attention on tcl/rvt
> test scripts
> that could be found in tests directory.
> First of all it needs to be tuned for apache 2 support, after that
> i'll need to  revue the scripts, probably  add a number of new ones.
> The purpose of this is to create some sort of automated suite which
> will allow us to test different aspects of rivet2 module. And will
> clear show up what does not work, and will make easy to catch up
> regressions.
> All thoughts on this topic are welcome.

My thoughts are basically "thankyou!".  Automated testing is a great
way of keeping a project free of bugs - every time you find a new one,
you can add a test that demonstrates it, and makes sure the bug won't
come back at some later date when someone says "this code is weird,
let's try getting rid of it and see what happens".

-- 
David N. Welton
http://www.welton.it/davidw/

---------------------------------------------------------------------
To unsubscribe, e-mail: rivet-dev-unsubscribe@tcl.apache.org
For additional commands, e-mail: rivet-dev-help@tcl.apache.org


Re: Planning

Posted by Massimo Manghi <ma...@unipr.it>.
David Welton wrote:
>> As I already told David, I'd like
>> if we started to think about a planning of the
>> project (as Valery suggested some weeks ago)
>> That would give me a chance to concentrate myself
>> on some aspects or modules and maybe be productive.
>>     
>
> What is there to plan, really?  Some coordination is necessary, such
> as "I'm going to be working on this file today", but aside from
> architectural questions, Rivet needs doing rather than planning,
> IMO:-)
>
>   
something like that is already a form of planning, but we should
do things to make rivet work also keeping the architectural
problems in mind (when possible): the more we can reuse in the
future the  better is.

 -- M

---------------------------------------------------------------------
To unsubscribe, e-mail: rivet-dev-unsubscribe@tcl.apache.org
For additional commands, e-mail: rivet-dev-help@tcl.apache.org


Re: Planning

Posted by David Welton <da...@gmail.com>.
> As I already told David, I'd like
> if we started to think about a planning of the
> project (as Valery suggested some weeks ago)
> That would give me a chance to concentrate myself
> on some aspects or modules and maybe be productive.

What is there to plan, really?  Some coordination is necessary, such
as "I'm going to be working on this file today", but aside from
architectural questions, Rivet needs doing rather than planning,
IMO:-)

-- 
David N. Welton
http://www.welton.it/davidw/

---------------------------------------------------------------------
To unsubscribe, e-mail: rivet-dev-unsubscribe@tcl.apache.org
For additional commands, e-mail: rivet-dev-help@tcl.apache.org