> 
> > The correct way to fix this whole situation is for
> > Linux developers to migrate to Solaris, and forget
> > about Linux. That would fix all these compilation
> > issues.
> 
> OOh, I like this one. Forget gcc compatibility. Kill
> Sun Studio gcc extension support now! Just make that
> hoard of gcc extension using developers port their
> code to Solaris for their users. Sorry? No users
> using
> Solaris? No demand? You get treated like dirt?


Actually, you both have a point.

First, I don't know that he was specifically referring to gcc extensions.

But it's IMO not wrong to complain about them; it's not quite as if
gcc and Studio are the only compilers in the universe, they're just
the main ones on Solaris (and of course gcc does have the point in
its favor that it's just about everywhere else, too).  In other words,
porting to Solaris may not be the only place where unencapsulated
dependence on gcc-isms causes problems.  (encapsulated use where needed,
so long as alternatives are generally available, isn't such a big deal,
as it can be much less of an obstacle to porting)

OTOH, it's a fact that there's a lot of that sort of code out there, and
I certainly don't have a problem with Studio picking up compatibility
features, as long as it retains a more strict mode of operation compatible
with existing Studio makefiles and existing Studio-compiled C++ object
files; the latter also because it's necessary to remain aware of when one
is using extensions, so that one can avoid them when appropriate.
Failing the realism of an extension-free environment (since there
are some cases where extensions are needed, although less than where
they are used), having two different sets of extensions at least allows
one to remain aware of such problem areas.

That highlights one of the problems that I have with some of what I
perceive happening with some open-source developers: they either
aren't aware when they're doing platform or compiler specific things,
or they just don't care, because they're more concerned with getting
something running on _their_ platform using _their_ first choice of
tools than with taking the time to familiarize themselves with portability
issues enough that the ability to port their work won't be just an afterthought.
Indeed, I suspect some of them would just as soon not have their work run
on anything but their preferred platform, which strikes me as more or
less contrary to the notion of open source, and going past open source 
pragmatism, past even license ideology, and off into platform religion.
 
 
This message posted from opensolaris.org
_______________________________________________
opensolaris-discuss mailing list
[email protected]

Reply via email to