On Tue, Sep 23, 2025 at 08:47:58AM +0200, Antoine Jacoutot wrote:
> On Sun, Sep 21, 2025 at 01:49:25PM +0200, Antoine Jacoutot wrote:
> > Hi.
> > 
> > This is a first tentative update diff for the latest webkitgtk 2.50.0 release.
> > Sending now so that it has a chance to make release and be able to get security
> > updates (webkitgtk does not do stable releases -1 nor backports, it just
> > increments).

Makes sense to me.

> > This is not the final diff, I need to see how to re-enable the GPU process
> > without dma-buf support which OpenBSD does not support.

:-/

> > There might also be a few adjustments here and there.
> > 
> > But that should be enough to test whether it builds and runs on !amd64
> > platforms. I am in the process of testing arm64.

FWIW all three flavors built for me on arm64 w/IBT.

> > If it fails to build on 32-bit platforms, let me know if this fixes it:
> > https://github.com/WebKit/WebKit/pull/51010
> > 
> > Thanks a lot!
> 
> New diff with fixed PLISTs.

As mentioned privately, something like this appears needed for
riscv64.  Build still ongoing.


--- patches/patch-Source_JavaScriptCore_assembler_MacroAssemblerRISCV64_h.orig	Sun Sep 28 12:15:31 2025
+++ patches/patch-Source_JavaScriptCore_assembler_MacroAssemblerRISCV64_h	Mon Sep 29 02:44:23 2025
@@ -1,3 +1,5 @@
+First hunk:
+
 https://github.com/WebKit/WebKit/pull/36459/
 
 From de57a167af8ec11f1f12e2d1f40bcd60334f77d6 Mon Sep 17 00:00:00 2001
@@ -12,6 +14,11 @@ Date: Mon, 11 Nov 2024 00:56:40 +0800
     * Source/JavaScriptCore/assembler/MacroAssemblerRISCV64.h:
     (JSC::MacroAssemblerRISCV64::add8):
 
+Second hunk:
+
+add missing implementation for convertUInt32ToDouble (and Float), used at
+least in Source/JavaScriptCore/bytecode/InlineCacheCompiler.cpp
+
 Index: Source/JavaScriptCore/assembler/MacroAssemblerRISCV64.h
 --- Source/JavaScriptCore/assembler/MacroAssemblerRISCV64.h.orig
 +++ Source/JavaScriptCore/assembler/MacroAssemblerRISCV64.h
@@ -30,3 +37,20 @@ Index: Source/JavaScriptCore/assembler/MacroAssemblerR
      void add32(RegisterID src, RegisterID dest)
      {
          add32(src, dest, dest);
+@@ -3544,6 +3552,16 @@ class MacroAssemblerRISCV64 : public AbstractMacroAsse
+     void convertInt32ToDouble(RegisterID src, FPRegisterID dest)
+     {
+         m_assembler.fcvtInsn<RISCV64Assembler::FCVTType::D, RISCV64Assembler::FCVTType::W>(dest, src);
++    }
++
++    void convertUInt32ToFloat(RegisterID src, FPRegisterID dest)
++    {
++        m_assembler.fcvtInsn<RISCV64Assembler::FCVTType::S, RISCV64Assembler::FCVTType::WU>(dest, src);
++    }
++
++    void convertUnt32ToDouble(RegisterID src, FPRegisterID dest)
++    {
++        m_assembler.fcvtInsn<RISCV64Assembler::FCVTType::D, RISCV64Assembler::FCVTType::WU>(dest, src);
+     }
+ 
+     void convertInt32ToDouble(TrustedImm32 imm, FPRegisterID dest)


-- 
jca