On Wed, Jul 19, 2023 at 03:59:59PM +0200, Antoine Jacoutot wrote: > On Wed, Jul 19, 2023 at 02:29:47PM +0200, Jeremie Courreges-Anglas wrote: > > > > Since upstream webkit(gtk) switched to JIT by default on riscv64, > > www/webkitgtk4 doesn't build on riscv64. > > > > > > http://build-failures.rhaalovely.net/riscv64/2023-06-30/www/webkitgtk4%2Cwebkitgtk41.log > > > > LowLevelInterpreter.asm doesn't seem fit for architectures that use the > > C_LOOP interpreter (powerpc, powerpc64, sparc64...) instead of JIT. > > Some fix might be needed here. > > > > Regarding riscv64, I suggest we go with the flow and enable the JIT on > > riscv64. The diff below drops one patch and adds another one: > > RISCV64Assembler.h assumes that intptr_t and int64_t are the same type, > > which isn't the case on OpenBSD (long vs long long) ; I merely added > > workarounds. Obviously MachineContext.h needs definitions for riscv64 > > too, I hope I haven't messed on of these. > > > > Using this diff, libexec/jsc computes 1+1 without crashing. Sadly I was > > not able to test libexec/MiniBrowser over SSH X11 forwarding (WebProcess > > appears to crash with an X11 error[0], just like on amd64). > > > > ok? > > Yup.
With a bump I assume?
