[phobos] Why we need to STOP breaking existing code

[Jesse Phillips]

I just tried the latest compiler for my project. In my own code I had
to change some code from deprecation.

As for DFL, it looks to all be due to bug fixes (haven't resolved the
issues so more may be hidden). In the past it has broken from const
changes.

form.d(493): Error: function
dfl.application.Application.addMessageFilter (IMessageFiltermf) is not
callable using argument types (const(FormMessageFilter))
form.d(2569): Error: function
dfl.application.Application.removeMessageFilter (IMessageFilter mf) is
not callable using argument types (const(FormMessageFilter))

Hmm, const variable can't be passed as non-const? Pretty sure that
should not have worked to begin with.

drawing.d(3630): Error: enum member DISPLAY not represented in final switch
splitter.d(264): Error: enum member NONE not represented in final switch
splitter.d(264): Error: enum member FILL not represented in final switch

Final switch not containing enum value? hmm, pretty sure that should
have been an error already.

So aside from not fixing bugs, it seems this was unavoidable. And note
in the email there was another complaint hidden there.

"Absolutely everything needs to be compiled before you can use it"

The linker will fail to link with the DFL libraries generated with dmd
2.059. So this would be a dmd thing, and I thought expected to some
degree (isn't this true with GCC on some version bumps?).

On Mon, Jul 30, 2012 at 4:35 PM, Walter Bright <walter at digitalmars.com> wrote:
> "DFL won't compile"
>
> It doesn't matter if the fix is trivial, easy, whatever. It does not
> compile, therefore D sux.