On Mon, 2012-05-28 at 12:34 +0300, Lauri Kasanen wrote: > On Mon, 28 May 2012 10:42:00 +0200 > Davidlohr Bueso <[email protected]> wrote: > > > > Surely you too see it's a road of diminishing gains? Would you like to > > > add code to support GCC 0.95? > > > > Care to explain why diminishing gains? If a new gcc version adds some > > amazing feature that only means older users won't take advantage of it, > > it won't affect them any other way. > > It is extra effort to do the workarounds for each older version, more so the > older they are. > > > There's a difference between totally stupid (0.95) and practical. > > Indeed. And I want that clarified and marked up. So you're not willing to > support 0.95; what is the lowest you _are_ willing to support? > > > You're mentioning pragmas and atomic builtins (that *already* broke > > current functionality for some platforms!!!) as examples. Pragmas are > > bogus and that's one of the reasons you see them avoided in many many > > projects (including the kernel - as much as possible). We got the exact > > same function with function attributes. > > And none of those platforms were mentioned by name. I find that rather, hm, > quirky, to say "it broke some holy platform whose name I may not mention". >
>From discussions with Eduardo, I believe it was Android that was broken. > The attributes do have the same effect, at the cost of an order or two > magnitudes of more writing and reading. We can spare the extra LoC - mk_macros is more than adequate for compiler detection and generic definitions. > > > Again, I ask you to please list all of those holy platforms and the versions > of GCC they carry, to constructively determine an acceptable minimum version. > 4.0.4 seems pertinent - January 31, 2007 (http://gcc.gnu.org/gcc-4.0/) That should be tolerant enough, I doubt that there'd be reports for broken platforms. - Davidlohr _______________________________________________ Monkey mailing list [email protected] http://lists.monkey-project.com/listinfo/monkey
