Re: [Firebird-test] What to do with TCS

[John B. <bel...@cs...>]

On Thursday, August 9, 2001, at 08:14 AM, Mark O'Donohue wrote:

> John Bellardo wrote:
>>
>> But we need to decide what to do with TCS first...
>
> Well Im all for giving it the flick, and extracting the tests into CVS 
> as flat files :-).
>
> But I'd really like to hear what Paul Reeves has to say (and probably 
> Frank), since they are the ones that have really done some worked with 
> it.  And I suspect they will have found some merrit in keeping it.

It does seem like a decent tool for handling the cross platform issues.  
The isql/qli scripts are not a big deal, but it does correctly handle 
c/c++ compilation which is a bonus.

>
> One think I have been thinking off, is a modified tcs program that 
> works from flat files could still be run from dejagnu, with perhaps an 
> option to use dejagnu to filter and compare the output.

Right, I was toying with that idea as well.  We could have a 
non-interactive mode for TCS where it would just do all the macro 
substitutions (and the like) it does now.  Then take the output script 
and run it with expect/tcl.  It would make a number of things that 
aren't working 100% in tcl work fine.  And we still have a TCS program 
to distribute to users (without the need for dejagnu) that they can use 
to perform some tests.

>
> TCS seems to me to represent a fair bit of effort in setting up and 
> running tests, but not as flexible at validating the generated output. 
> The tcl/expect framework, is pretty good at that, so perhaps a 
> combination could be used, TCS to run these tests and dejagnu to do the 
> compare of the output.

The biggest problem with output compare is expect merges the 
stdout/stderr streams, while TCS puts all stderr output first, followed 
by all stdout output.  I've already dealt with this is a great extent in 
the current TCS/Tcl, but I'm sure there are still issues.

-John