Re: [MiNT] BubbleGEM

[Jo Even Skarstein <joska@online.no>]

On Sat, 2011-01-08 at 13:23 +0100, Peter Persson wrote:

> I'm not familiar with the inner workings of BubbleGEM. Is it a kludge because the protocol/API itself is bad, or because of a bad implementation?

I have not studied the BubbleGEM sources, so I know little about the
implementation. I do know that the CPX crashes with an address error
with MP enabled though, and that's a good reason for not using it.

The problem is how applications calls BubbleGEM. You send it a message
with a pointer to the string. BubbleGEM then copies the string and
returns an ack-message. Implementing tooltip-functionality in the kernel
with such a protocol would be madness.

You could do it properly and add an AES call for the
tooltip-functionality. This would be clean, and much simpler for the
calling application. However, it would be of limited use as all existing
apps that use tooltips expect the BubbleGEM protocol. You would have to
write a wrapper to allow older apps to use the tooltips, and then you've
gained nothing.

The BubbleGEM protocol doesn't require any dirty tricks to work. Passing
pointers in AES messages is questionable, but OK when done properly. So
making a clean, non-blocking (using windows) BubbleGEM-clone is probably
the easiest and most flexible solution.

Jo Even