Re: riscv64 support for lang/node

On 2022/09/01 20:56, Volker Schlecht wrote:
> Can't test on riscv64 nor OK anything, but it applys cleanly to my node

as a port maintainer you can OK things for that port..

> 16.17.0 patch that's still waiting for a go ... **hint**
>
> Combined diff attached if you'd have a look ;-)
>
> On 9/1/22 02:50, Jeremie Courreges-Anglas wrote:
> >
> > Hopefully the comments in the patches explain what is going on. No
> > REVISION bump since all the patched are riscv64-specific or (AFAIK)
> > not affecting any arch where node currently builds
> > (patches/patch-tools_v8_gypfiles_v8_gyp). But I can add a REVISION bump
> > if it feels more comfortable.
> >
> > With this I can attempt building firefox (build ongoing) or chromium.
> >
> > ok?
> >
> >
> > Index: Makefile
> > ===================================================================
> > RCS file: /cvs/ports/lang/node/Makefile,v
> > retrieving revision 1.105
> > diff -u -p -r1.105 Makefile
> > --- Makefile 22 Jul 2022 17:46:49 -0000 1.105
> > +++ Makefile 30 Aug 2022 15:44:48 -0000
> > @@ -1,4 +1,4 @@
> > -ONLY_FOR_ARCHS = amd64 aarch64 i386 powerpc64
> > +ONLY_FOR_ARCHS = amd64 aarch64 i386 powerpc64 riscv64
> > DPB_PROPERTIES = parallel
> > USE_WXNEEDED = Yes
> > Index: patches/patch-deps_v8_src_codegen_riscv64_assembler-riscv64_cc
> > ===================================================================
> > RCS file: patches/patch-deps_v8_src_codegen_riscv64_assembler-riscv64_cc
> > diff -N patches/patch-deps_v8_src_codegen_riscv64_assembler-riscv64_cc
> > --- /dev/null 1 Jan 1970 00:00:00 -0000
> > +++ patches/patch-deps_v8_src_codegen_riscv64_assembler-riscv64_cc 31 Aug 2022 09:04:56 -0000
> > @@ -0,0 +1,17 @@
> > +Use static_cast instead of a potentially dangerous reinterpret_cast.
> > +Doesn't matter since this file only supports riscv64 where unsigned long and
> > +unsigned long long are actually the same size.
> > +../deps/v8/src/codegen/riscv64/assembler-riscv64.cc:403:22: error: reinterpret_cast from 'v8::internal::Address' (aka 'unsigned long') to 'uint64_t' (aka 'unsigned long long') is not allowed
> > +
> > +Index: deps/v8/src/codegen/riscv64/assembler-riscv64.cc
> > +--- deps/v8/src/codegen/riscv64/assembler-riscv64.cc.orig
> > ++++ deps/v8/src/codegen/riscv64/assembler-riscv64.cc
> > +@@ -400,7 +400,7 @@ int Assembler::target_at(int pos, bool is_internal) {
> > + Address pc = reinterpret_cast<Address>(buffer_start_ + pos);
> > + pc = target_address_at(pc);
> > + uint64_t instr_address = reinterpret_cast<uint64_t>(buffer_start_ + pos);
> > +- uint64_t imm = reinterpret_cast<uint64_t>(pc);
> > ++ uint64_t imm = static_cast<uint64_t>(pc);
> > + if (imm == kEndOfJumpChain) {
> > + return kEndOfChain;
> > + } else {
> > Index: patches/patch-deps_v8_src_codegen_riscv64_cpu-riscv64_cc
> > ===================================================================
> > RCS file: patches/patch-deps_v8_src_codegen_riscv64_cpu-riscv64_cc
> > diff -N patches/patch-deps_v8_src_codegen_riscv64_cpu-riscv64_cc
> > --- /dev/null 1 Jan 1970 00:00:00 -0000
> > +++ patches/patch-deps_v8_src_codegen_riscv64_cpu-riscv64_cc 1 Sep 2022 00:28:54 -0000
> > @@ -0,0 +1,39 @@
> > +Implement instruction cache flush using sysarch(RISCV_SYNC_ICACHE).
> > +
> > +Index: deps/v8/src/codegen/riscv64/cpu-riscv64.cc
> > +--- deps/v8/src/codegen/riscv64/cpu-riscv64.cc.orig
> > ++++ deps/v8/src/codegen/riscv64/cpu-riscv64.cc
> > +@@ -4,7 +4,13 @@
> > +
> > + // CPU specific code for arm independent of OS goes here.
> > +
> > +-#include <sys/syscall.h>
> > ++#ifdef __OpenBSD__
> > ++# include <sys/types.h>
> > ++# include <machine/sysarch.h>
> > ++#else
> > ++# include <sys/syscall.h>
> > ++