__sa is an old magic to specify the size of gc_header_s, I removed it in the recent commit.
For newlisp, I noticed it before. I think one of the difference is that newlisp is trying to provide a completed I/O routines set, while SEE is not. SEE may be considered minimalistic :) And there is no x86_64 environment for me :(, I may try to create a x86_64 VM to fix the potential bugs in the code. On Sun, Oct 23, 2011 at 7:44 PM, pancake <panc...@youterm.com> wrote: > More notes: > - add a space between -o and the filename when linking. This is also breaking > the build. > > - defining gcheader foo to 24 makes the build happen, but full of pointer/int > warnings which i imagine this is just wrong pointer arithmetics that will > result in segfaults. > > - im also seeing wrong format strings like using %x instrad of %p .. > > - make clean fails > > Im not going to execute something that the compilers says its wrong at first > place. The idea looks cool but needs a big cleanup in order to make it happy > with pointers and be 64bit friendly. > > - i would like it more if the LOCs get under 2kloc :P > > On 23/10/2011, at 12:30, Pierre Chapuis <catw...@archlinux.us> wrote: > >> On 23.10.2011 12:20, Pierre Chapuis wrote: >> >>> Precision: breaks *on a 64 bits machine* (gc_header_t is a pointer >>> so gc_header_s is 8 bytes larger than GC_HEADER_SPACE). >> >> OK, apparently just setting GC_HEADER_SPACE to 24 makes it build >> and run simple tests. Sorry for the noise. >> > >
