Re: [Opencxx-users] testsuit

[<se...@in...>]

On Tue, 2004-08-10 at 19:16, Stefan Seefeld wrote:
> Gilles J. Seguin wrote:
> 
> > Tests for the occ binary on the PATH environment variable
> >  <http://www.infonet.ca/segg/occ.tgz>
> > will unpack in occ directory.
> > $ cd occ
> > $ qmtest run	# will run everything.
> > $ qmtest run parse  # will do the same as 'qmtest run'
> > since we have only one directory.
> > $ qmtest run parse.access1  # will run only one test.
> > 
> > The tests examine results from 'occ -p -s myfile.cc'
> 
> nice ! Some suggestions, though:
> 
> you hardcode the path '/usr/bin/occ' in each single test descriptor.
> I'v just changed one to see that it works, and I'm not going to
> edit all these files :-)
> 
> What I'd suggest to do instead is defining a new test (python) class
> that is then registered with the database and used for all tests.
> Then you can configure paths and other stuff right into this class
> *once*.
> 
> Second I think it would be good to substructure the tests. In fact
> I'd suggest to write a new 'test database', and then use the subdivision
> into test suites to set up test types for each suite separately.

tests are stolen from gcc/gcc/testsuite/g++.dg with branch tag
gcc-3_4-rhl-branch.
And follow the same structure.
The tar file was a copy of a test directory executing occ on
uninstall binary and library.  On which I have apply sed command on
all file.

> That will allow to tailor the test types, for example one suite/folder
> to test the Encoding class, one to test ptree generation / manipulation,
> etc., etc.

What I am trying to do now, is to generate source files base on
rules(rProgram, rDeclaration, rClassSpec).  Are they producing
valid ASTs ?
First step, document the Grammer. 

> Note that this is exactly the approach I'v taken in synopsis, i.e.
> I wrote my own database and a handfull of test classes to do various
> things (one test class compiles a C++ file and executes it, another
> runs a particular synopsis processor on a specific input file, you
> get the idea...)
> 
> That said, I think what you'v done is a nice starting point ! May
> I integrate it into my existing testing harness ?
> 
> Regards,
> 		Stefan