I thought of a simple way to do the multi-lib thing for the moment(phobos,etc.).  Add a compiler switch to define which phobos.lib to link. -p or something like that.  If it wasn't there, it would just revert to the default behavior if the switch wasn't there.  This would make it a lot simpler to have several versions of phobos in use.

A switch to replace snn.lib could also be an option, if that library is open source.  If you want it to be opaque, it would be nice for you to define what things it provides to D, so someone writing another implementation would know what it does.

Another thing that would be nice is to have better documentation on how exactly the GC is interfaced  with D itself, so one it'd be easier to completely replace, should someone feel the need(where new/delete are defined, what exactly they call, how and when the scanning is done, etc.).  Anyway, let me know what you think.

Evan

"Evan McClanahan" <evan@dontSPAMaltarinteractive.com> schrieb im Newsbeitrag news:ar88aa$i18$1@digitaldaemon.com...

> I thought of a simple way to do the multi-lib thing for the moment(phobos,etc.).  Add a compiler switch to define which phobos.lib to link. -p or something like that.  If it wasn't there, it would just revert to the default behavior if the switch wasn't there.  This would make it a lot simpler to have several versions of phobos in use.

HI, I like the idea! It makes life & experiments a lot easier.

> A switch to replace snn.lib could also be an option, if that library is open source.  If you want it to be opaque, it would be nice for you to define what things it provides to D, so someone writing another implementation would know what it does.

I second this one too.

> Another thing that would be nice is to have better documentation on how exactly the GC is interfaced  with D itself, so one it'd be easier to completely replace, should someone feel the need(where new/delete are defined, what exactly they call, how and when the scanning is done, etc.).  Anyway, let me know what you think.

And again I agree. IMO it would be greate value to D if the GC could be switched. First you could use GCs with different strategies depening on your application. And further it enables us to do our own research in this area without having to implement a complete compiler. Robert

"Robert M. Münch" <robert.muench@robertmuench.de> wrote in message news:ara3oj$29jc$1@digitaldaemon.com...
> And again I agree. IMO it would be greate value to D if the GC could be switched. First you could use GCs with different strategies depening on
your
> application. And further it enables us to do our own research in this area without having to implement a complete compiler. Robert

The gc code is separated out in phobos; phobos itself incorporates dmgc.lib. You can write your own gc and insert it if you want. -Walter