On 5 November 2012 18:04, H. S. Teoh <hst...@quickfur.ath.cx> wrote: > On Mon, Nov 05, 2012 at 05:40:36PM +0000, Iain Buclaw wrote: >> On 5 November 2012 16:10, Joseph Rushton Wakeling >> <joseph.wakel...@webdrake.net> wrote: > [...] >> > Is there a particular reason not to allow the frontend, druntime and >> > phobos to be git submodules which could be kept up to date with >> > upstream in a more 'live' fashion? I can imagine it might add extra >> > day-to-day issues but it might save the pain of those huge merges >> > ... >> > >> >> Because we make changes to DFE, Phobos, Druntime to be compatible with >> things GDC does that DMD doesn't do. > > If we somehow manage to get 2.060 working for the 4.7 branch, would you > take a pull request? ;-) > > I know you're focusing on 4.8, but I'm still unable to get it to build > correctly, whereas the 4.7 branch works for me. And I'd really like to > have 2.060 gdc for some of my D projects. > >
Yes. >> I have discussed the notion of altering the DFE to be a more shared >> source base for all compilers - where areas that are DMD-specific in >> DFE are removed and replaced with hooks. So to port DFE to your own >> backend, all you require to do is implement these hooks in your glue >> code (on top of all the ::toSymbol, ::toIR, ::toObjFile, ::toCtype, >> routines). > [...] > > I think this is something important to work towards. This will greatly > widen the availability of D, which can only be a good thing. > > Andrei thought very much the same when I last spoke to him, but it's later down the priority list - eg, ARM is a bigger priority atm. -- Iain Buclaw *(p < e ? p++ : p) = (c & 0x0f) + '0';
