Re: [Cppunit-devel] CppUnit 1.7.10 tar ball

[Baptiste L. <gai...@fr...>]

----- Original Message -----
From: "Robert Wenner" <ro...@po...>
To: "Baptiste Lepilleur" <gai...@fr...>
Sent: Thursday, April 04, 2002 12:56 PM
Subject: Re: [Cppunit-devel] CppUnit 1.7.10 tar ball


> Hi!
>
> Sorry if you get this email more than once, it was returned yesterday
> because it could not be delivered for 4 hours. There seems to be a
> problem with my mail going to the list.
>
> --- cut here ---
>
> Latest version seems to be almost fine under Linux for me, too.
>
> During installation I get these messages:
> $ autoconf
> You should update your `aclocal.m4' by running aclocal.


Don't know anything about that. Somebody does ?

> And later:
> $ make install
> [...]
> /usr/bin/install: cannot create regular file
> `/usr/local/share/cppunit/html/html/index.html': No such file or
> directory
> make[2]: *** [install-htmlDATA] Error 1
> That's one "html" too much?

You may not have installed doxygen (the documentation generator). It can be
found at:
http://www.doxygen.org/. Though, that is kind of weird since we provide the
generated documentation in a separate bundle. That lead me to beleive this
would be a bug in the makefile (don't know how to fix that one).

>
> On using the deprecated TextTestRunner.h the preprocessor could not find
> the include file, I changed in include/cppunit/TextTestRunner.h this
> line
> #include <cppunitui/text/TestRunner.h>
> to
> #include <../cppunitui/text/TestRunner.h>

Hum, something is wrong there. The include hierarchy is supposed to be like
this:

include/
    cppunit/
        TextTestRunner.h
    cppunitui/
        text/
            TestRunner.h

Since include/ is the directory given to the compiler to search include, it
should be resolved successfully (and was on VC++ before I mad the update in
the examples). Could you check that include/ is the directory you give to
the compiler ?

    An acceptable fix would be:
#include "../cppunitui/text/TestRunner.h"
    => the include is relative to the locate of TextTestRunner.h file, while
#include <...> just pick any directory given to the compiler which resolved
sucessfully the specified path. Since CppUnit only require include/, the
sucessful resolution can only appear with a "miraculous" other unspecified
directory...

  Let's me know how it goes.

>
> Documentation is somewhat better now. :-)
>
> Anyway I still don't like the CppUnit::TextUi::TestRunner
> implementation. I don't like the call to runner.run with its boolean
> arguments for progress printing, waiting and summary printing. We do
> have listeners, if I do not add a listener I obvisiously do not want
> progress printing or waiting. If I do not add an outputter I
> obvisiously do not want a summary.
> I think the runner should never ever add a listener on its own,
> especially not like in TestRunner.cpp in method runTest (Test*, bool)
> around line 131.
> IMHO that's a good place to do some refactoring.

I'll probably fix that after 1.8.0. I think I'll introduce a
CppUnit::TestRunner class which does not use any outputter, but create a
default TestResult and give access to the TestResult. run() would run the
specified test and that's it.


>
> Another place for refactoring: have the CPPUNIT_ASSERT_EQUAL_MESSAGE
> macro take the message as its first argument as CPPUNIT_ASSERT_MESSAGE
> macro does. (Or make the message always the last argument, but make it
> always the same way.)

I made the change in the current CVS version. message is now the first
argument.

Thanks for the feedback,
Baptiste.