[csw-maintainers] MySQL built with GCC → mysql config vs bindings

[Peter FELECAN]

"Maciej (Matchek) Bliziński" <maciej at opencsw.org> writes:

> 2012/6/19 Peter FELECAN <pfelecan at opencsw.org>:
>> Ben Walton <bwalton at opencsw.org> writes:
>>
>>> Unless we make a hard standardization on one compiler, this problem
>>> won't be easily solved, I don't think.
>>>
>>> What do others think?
>>
>> For the majority of projects coming from the GNU and associates, using
>> gcc is the best thing that we can do. To simplify thing, all libraries
>> and development tools should be built with gcc; as a side effect we
>> avoid /opt/csw/gxx and other horrors. Using studio can be an option for
>> end user applications, i.e. on which there are no dependencies; of
>> course, for those application written in C++ it becomes mandatory to use
>> gcc.
>>
>> These were my 2 drachmas of evening wisdom
>
> I wouldn't like this to go unanswered. In my opinion, this is a
> reasonable target state. I don't mean that there should be only one
> compiler in the world. If C and C++ are portable languages, there
> should be multiple compilers implementing standards and code should be
> good to compile with any of them, on any system. Getting the state
> where the code compiles with multiple compilers requires a lot of
> work. Do we want us to be the people who do that? Many upstream
> developers don't test on Solaris, which already creates a lot of work
> of us, maintainers. Adding one more dimension of difference, we're
> putting even more work on our plate. Not that porting isn't
> interesting ‒ when building with Studio, you get to learn about some
> interesting aspects of C and C++. But with time, if what you want is a
> working binary and not porting fun, it just gets tedious.
>
> I suspect that we wouldn't be able to maintain Solaris Studio ports of
> all the software that we maintain. Some software we just want built
> and released. At the same time, I see value in building at least some
> major software projects, such as the database engines, with more than
> just GCC. If they already do a good job of building with Studio, why
> not continue building them with Studio? I just wouldn't like the
> compiler porting issue to hamper our ability to keep up to date with
> upstream releases and new builds.

Here are 2 recent cases:

- packaging pilot-link I'm forced to use Solaris Studio because TCL is
  built with and I'm left with the choice between compiling with gcc but
  without TCL binding or with Solaris Studio and TCL binding but
  probably forced to use the same for JPilot... I choose to go with the
  TCL binding as I'm taking over the package and don't wish to offer a
  poorer experience.

- psedit has partially C++ code but depending on packages built with
  Solaris Studio, I'm forced to use it; this is not a case of a better
  complying compiler but of different mangling algorithms, isn't it?

Given our limited resources we should be Hamiltonian adepts and adopt
policies in this direction, i.e., using gcc as the default compiler.
-- 
Peter