svn commit: r383466 - in head/x11-toolkits/wxgtk30: . files

[Alexey Dokuchaev]

On Sun, Apr 12, 2015 at 01:24:01PM -0700, Don Lewis wrote:
> On 12 Apr, Alexey Dokuchaev wrote:
> > On Sun, Apr 12, 2015 at 03:09:00PM +0300, Dmitry Marakasov wrote:
> >> * John Marino (freebsd.contact at marino.st) wrote:
> >> > > [...]
> >> > > A prime example of why c++ sucks by the way.
> >> 
> >> Nothing like that. It's just that crappy incompatible code sucks,
> >> instead we could just use c++11 by default everywhere, which would
> >> automatically get us some performance gains.
> > 
> > Of course it sucks.  Do you know another comparably prominent example
> > when things breaking so spectacularly all over the place? :-)
> > 
> > If one must use C++ to develop e.g. a library intended to be used in
> > open-source projects, they really should stick to c++03 (and make sure
> > their code is fine by gcc42) for at least couple of years from now.
> 
> Our copy of gcc42 in base doesn't claim to support anything newer than
> c++98.  This is what I get when I print ${COMPILER_FEATURES}:
> 	libstdc++ c89 c99 gnu89 gnu99 c++98 gnu++98

Certainly if you can stay within c++98 then you should.  On a larger scale,
gcc42 support is optional, so c++03 is pretty close to being a safe ground.

Striving to support C++ by gcc3.x and gcc2.95 is worthless; but your C code
should benefit from being gcc2.95-compatible.

Of course, more different compilers help portability.

These are basic, obvious observations about how to develop good and friendly
code in FOSS world.  I am honestly surprised when folks do not these things.

./danfe