[bug#63069] [PATCH] Patch `openjdk@9` so it builds on aarch64

[bug#63069] [PATCH] Patch `openjdk@9` so it builds on aarch64

[burning2007]

From: Anthony <burning2007@ya.ru>

Patch openjdk9 so it can build on aarch64. Adapted from here:

- https://bugs.openjdk.org/browse/JDK-8224851
- https://github.com/openjdk/jdk/commit/a0a919603c1546ce86fe659df3636b5a24b9caeb?diff=unified

I can confirm it builds on real hardware.

--
Anthony
---
 gnu/packages/java.scm                         |   3 +
 ...19603c1546ce86fe659df3636b5a24b9caeb.patch | 371 ++++++++++++++++++
 2 files changed, 374 insertions(+)
 create mode 100644 gnu/packages/patches/opendjk-9-aarch64-a0a919603c1546ce86fe659df3636b5a24b9caeb.patch

diff --git a/gnu/packages/java.scm b/gnu/packages/java.scm
index de3a0003e5..d389b5a902 100644
--- a/gnu/packages/java.scm
+++ b/gnu/packages/java.scm
@@ -856,6 +856,9 @@ (define-public openjdk9
               (uri (hg-reference (url "https://hg.openjdk.org/jdk/jdk")
                                  (changeset "jdk-9+181")))
               (file-name (hg-file-name name version))
+              (patches
+                (search-patches
+                  "opendjk-9-aarch64-a0a919603c1546ce86fe659df3636b5a24b9caeb.patch"))
               (modules '((guix build utils)))
               (snippet `(begin
                           (for-each delete-file
diff --git a/gnu/packages/patches/opendjk-9-aarch64-a0a919603c1546ce86fe659df3636b5a24b9caeb.patch b/gnu/packages/patches/opendjk-9-aarch64-a0a919603c1546ce86fe659df3636b5a24b9caeb.patch
new file mode 100644
index 0000000000..c55a3253ee
--- /dev/null
+++ b/gnu/packages/patches/opendjk-9-aarch64-a0a919603c1546ce86fe659df3636b5a24b9caeb.patch
@@ -0,0 +1,371 @@
+From a0a919603c1546ce86fe659df3636b5a24b9caeb Mon Sep 17 00:00:00 2001
+From: Nick Gasson <ngasson@openjdk.org>
+Date: Mon, 17 Jun 2019 15:31:49 +0800
+Subject: [PATCH] 8224851: AArch64: fix warnings and errors with Clang and GCC
+ 8.3
+
+Reviewed-by: aph, kbarrett
+---
+ hotspot/src/cpu/aarch64/vm/aarch64.ad            |  8 +--
+ hotspot/src/cpu/aarch64/vm/assembler_aarch64.hpp |  4 +-
+ .../cpu/aarch64/c1_LIRAssembler_aarch64.cpp   | 12 ++--
+ .../cpu/aarch64/c1_LIRGenerator_aarch64.cpp   |  4 +-
+ hotspot/src/cpu/aarch64/vm/frame_aarch64.cpp     | 10 +--
+ .../cpu/aarch64/interp_masm_aarch64.hpp       |  2 -
+ .../cpu/aarch64/macroAssembler_aarch64.cpp    |  2 +-
+ .../aarch64/macroAssembler_aarch64_log.cpp    |  4 +-
+ .../aarch64/macroAssembler_aarch64_trig.cpp   | 30 ++++-----
+ .../cpu/aarch64/vm_version_aarch64.cpp        |  2 +-
+ .../linux_aarch64/atomic_linux_aarch64.hpp    |  4 +-
+ .../os_cpu/linux_aarch64/copy_linux_aarch64.s |  2 +-
+ .../os_cpu/linux_aarch64/os_linux_aarch64.cpp | 61 ++++++-------------
+ 13 files changed, 62 insertions(+), 83 deletions(-)
+
+diff --git a/hotspot/src/cpu/aarch64/vm/aarch64.ad b/hotspot/src/cpu/aarch64/vm/aarch64.ad
+index e6a1bc8bba98..a5afb2a08fef 100644
+--- a/hotspot/src/cpu/aarch64/vm/aarch64.ad
++++ b/hotspot/src/cpu/aarch64/vm/aarch64.ad
+@@ -14255,7 +14255,7 @@ instruct compF_reg_zero(rFlagsReg cr, vRegF src1, immF0 src2)
+   format %{ "fcmps $src1, 0.0" %}
+ 
+   ins_encode %{
+-    __ fcmps(as_FloatRegister($src1$$reg), 0.0D);
++    __ fcmps(as_FloatRegister($src1$$reg), 0.0);
+   %}
+ 
+   ins_pipe(pipe_class_compare);
+@@ -14284,7 +14284,7 @@ instruct compD_reg_zero(rFlagsReg cr, vRegD src1, immD0 src2)
+   format %{ "fcmpd $src1, 0.0" %}
+ 
+   ins_encode %{
+-    __ fcmpd(as_FloatRegister($src1$$reg), 0.0D);
++    __ fcmpd(as_FloatRegister($src1$$reg), 0.0);
+   %}
+ 
+   ins_pipe(pipe_class_compare);
+@@ -14360,7 +14360,7 @@ instruct compF3_reg_immF0(iRegINoSp dst, vRegF src1, immF0 zero, rFlagsReg cr)
+     Label done;
+     FloatRegister s1 = as_FloatRegister($src1$$reg);
+     Register d = as_Register($dst$$reg);
+-    __ fcmps(s1, 0.0D);
++    __ fcmps(s1, 0.0);
+     // installs 0 if EQ else -1
+     __ csinvw(d, zr, zr, Assembler::EQ);
+     // keeps -1 if less or unordered else installs 1
+@@ -14387,7 +14387,7 @@ instruct compD3_reg_immD0(iRegINoSp dst, vRegD src1, immD0 zero, rFlagsReg cr)
+     Label done;
+     FloatRegister s1 = as_FloatRegister($src1$$reg);
+     Register d = as_Register($dst$$reg);
+-    __ fcmpd(s1, 0.0D);
++    __ fcmpd(s1, 0.0);
+     // installs 0 if EQ else -1
+     __ csinvw(d, zr, zr, Assembler::EQ);
+     // keeps -1 if less or unordered else installs 1
+diff --git a/hotspot/src/cpu/aarch64/vm/assembler_aarch64.hpp b/hotspot/src/cpu/aarch64/vm/assembler_aarch64.hpp
+index 3bb620d52e0a..a95417a467a9 100644
+--- a/hotspot/src/cpu/aarch64/vm/assembler_aarch64.hpp
++++ b/hotspot/src/cpu/aarch64/vm/assembler_aarch64.hpp
+@@ -276,7 +276,7 @@ class Instruction_aarch64 {
+   unsigned get(int msb = 31, int lsb = 0) {
+     int nbits = msb - lsb + 1;
+     unsigned mask = ((1U << nbits) - 1) << lsb;
+-    assert_cond(bits & mask == mask);
++    assert_cond((bits & mask) == mask);
+     return (insn & mask) >> lsb;
+   }
+ 
+@@ -2644,7 +2644,7 @@ void mvnw(Register Rd, Register Rm,
+   // RBIT only allows T8B and T16B but encodes them oddly.  Argh...
+   void rbit(FloatRegister Vd, SIMD_Arrangement T, FloatRegister Vn) {
+     assert((ASSERTION), MSG);
+-    _rbit(Vd, SIMD_Arrangement(T & 1 | 0b010), Vn);
++    _rbit(Vd, SIMD_Arrangement((T & 1) | 0b010), Vn);
+   }
+ #undef ASSERTION
+ 
+diff --git a/hotspot/src/cpu/aarch64/vm/c1_LIRAssembler_aarch64.cpp b/hotspot/src/cpu/aarch64/vm/c1_LIRAssembler_aarch64.cpp
+index 100b9e9c1154..eb013eb52be4 100644
+--- a/hotspot/src/cpu/aarch64/vm/c1_LIRAssembler_aarch64.cpp
++++ b/hotspot/src/cpu/aarch64/vm/c1_LIRAssembler_aarch64.cpp
+@@ -1078,8 +1078,8 @@ void LIR_Assembler::emit_opBranch(LIR_OpBranch* op) {
+       // Assembler::EQ does not permit unordered branches, so we add
+       // another branch here.  Likewise, Assembler::NE does not permit
+       // ordered branches.
+-      if (is_unordered && op->cond() == lir_cond_equal
+-          || !is_unordered && op->cond() == lir_cond_notEqual)
++      if ((is_unordered && op->cond() == lir_cond_equal)
++          || (!is_unordered && op->cond() == lir_cond_notEqual))
+         __ br(Assembler::VS, *(op->ublock()->label()));
+       switch(op->cond()) {
+       case lir_cond_equal:        acond = Assembler::EQ; break;
+@@ -1789,18 +1789,22 @@ void LIR_Assembler::arith_op(LIR_Code code, LIR_Opr left, LIR_Opr right, LIR_Opr
+     switch (code) {
+     case lir_add: __ fadds (dest->as_float_reg(), left->as_float_reg(), right->as_float_reg()); break;
+     case lir_sub: __ fsubs (dest->as_float_reg(), left->as_float_reg(), right->as_float_reg()); break;
++    case lir_mul_strictfp: // fall through
+     case lir_mul: __ fmuls (dest->as_float_reg(), left->as_float_reg(), right->as_float_reg()); break;
++    case lir_div_strictfp: // fall through
+     case lir_div: __ fdivs (dest->as_float_reg(), left->as_float_reg(), right->as_float_reg()); break;
+     default:
+       ShouldNotReachHere();
+     }
+   } else if (left->is_double_fpu()) {
+     if (right->is_double_fpu()) {
+-      // cpu register - cpu register
++      // fpu register - fpu register
+       switch (code) {
+       case lir_add: __ faddd (dest->as_double_reg(), left->as_double_reg(), right->as_double_reg()); break;
+       case lir_sub: __ fsubd (dest->as_double_reg(), left->as_double_reg(), right->as_double_reg()); break;
++      case lir_mul_strictfp: // fall through
+       case lir_mul: __ fmuld (dest->as_double_reg(), left->as_double_reg(), right->as_double_reg()); break;
++      case lir_div_strictfp: // fall through
+       case lir_div: __ fdivd (dest->as_double_reg(), left->as_double_reg(), right->as_double_reg()); break;
+       default:
+         ShouldNotReachHere();
+diff --git a/hotspot/src/cpu/aarch64/vm/c1_LIRGenerator_aarch64.cpp b/hotspot/src/cpu/aarch64/vm/c1_LIRGenerator_aarch64.cpp
+index 19c9ea688f6d..185e93b99f93 100644
+--- a/hotspot/src/cpu/aarch64/vm/c1_LIRGenerator_aarch64.cpp
++++ b/hotspot/src/cpu/aarch64/vm/c1_LIRGenerator_aarch64.cpp
+@@ -426,7 +426,7 @@ void LIRGenerator::do_ArithmeticOp_FPU(ArithmeticOp* x) {
+     tmp = new_register(T_DOUBLE);
+   }
+ 
+-  arithmetic_op_fpu(x->op(), reg, left.result(), right.result(), NULL);
++  arithmetic_op_fpu(x->op(), reg, left.result(), right.result(), x->is_strictfp());
+ 
+   set_result(x, round_item(reg));
+ }
+diff --git a/hotspot/src/cpu/aarch64/vm/frame_aarch64.cpp b/hotspot/src/cpu/aarch64/vm/frame_aarch64.cpp
+index 08b8f2b915b7..3019cb01acc6 100644
+--- a/hotspot/src/cpu/aarch64/vm/frame_aarch64.cpp
++++ b/hotspot/src/cpu/aarch64/vm/frame_aarch64.cpp
+@@ -768,11 +768,13 @@ extern "C" void npf() {
+ 
+ extern "C" void pf(unsigned long sp, unsigned long fp, unsigned long pc,
+                    unsigned long bcx, unsigned long thread) {
+-  RegisterMap map((JavaThread*)thread, false);
+   if (!reg_map) {
+-    reg_map = (RegisterMap*)os::malloc(sizeof map, mtNone);
++    reg_map = NEW_C_HEAP_OBJ(RegisterMap, mtNone);
++    ::new (reg_map) RegisterMap((JavaThread*)thread, false);
++  } else {
++    *reg_map = RegisterMap((JavaThread*)thread, false);
+   }
+-  memcpy(reg_map, &map, sizeof map);
++
+   {
+     CodeBlob *cb = CodeCache::find_blob((address)pc);
+     if (cb && cb->frame_size())
+diff --git a/hotspot/src/cpu/aarch64/vm/interp_masm_aarch64.hpp b/hotspot/src/cpu/aarch64/vm/interp_masm_aarch64.hpp
+index 72ec4cd1e36b..8168b50ec9f1 100644
+--- a/hotspot/src/cpu/aarch64/vm/interp_masm_aarch64.hpp
++++ b/hotspot/src/cpu/aarch64/vm/interp_masm_aarch64.hpp
+@@ -38,8 +38,6 @@ class InterpreterMacroAssembler: public MacroAssembler {
+  protected:
+ 
+  protected:
+-  using MacroAssembler::call_VM_leaf_base;
+-
+   // Interpreter specific version of call_VM_base
+   using MacroAssembler::call_VM_leaf_base;
+ 
+diff --git a/hotspot/src/cpu/aarch64/vm/macroAssembler_aarch64.cpp b/hotspot/src/cpu/aarch64/vm/macroAssembler_aarch64.cpp
+index 5bfa8133d5bb..1c264b4df248 100644
+--- a/hotspot/src/cpu/aarch64/vm/macroAssembler_aarch64.cpp
++++ b/hotspot/src/cpu/aarch64/vm/macroAssembler_aarch64.cpp
+@@ -2681,7 +2681,7 @@ Address MacroAssembler::spill_address(int size, int offset, Register tmp)
+   if ((offset & (size-1)) && offset >= (1<<8)) {
+     add(tmp, base, offset & ((1<<12)-1));
+     base = tmp;
+-    offset &= -1<<12;
++    offset &= -1u<<12;
+   }
+ 
+   if (offset >= (1<<12) * size) {
+diff --git a/hotspot/src/cpu/aarch64/vm/vm_version_aarch64.cpp b/hotspot/src/cpu/aarch64/vm/vm_version_aarch64.cpp
+index b917636cb635..dbf1dcdb5409 100644
+--- a/hotspot/src/cpu/aarch64/vm/vm_version_aarch64.cpp
++++ b/hotspot/src/cpu/aarch64/vm/vm_version_aarch64.cpp
+@@ -177,7 +177,7 @@ void VM_Version::get_processor_features() {
+   if (FILE *f = fopen("/proc/cpuinfo", "r")) {
+     char buf[128], *p;
+     while (fgets(buf, sizeof (buf), f) != NULL) {
+-      if (p = strchr(buf, ':')) {
++      if ((p = strchr(buf, ':')) != NULL) {
+         long v = strtol(p+1, NULL, 0);
+         if (strncmp(buf, "CPU implementer", sizeof "CPU implementer" - 1) == 0) {
+           _cpu = v;
+diff --git a/hotspot/src/os_cpu/linux_aarch64/vm/copy_linux_aarch64.s b/hotspot/src/os_cpu/linux_aarch64/vm/copy_linux_aarch64.s
+index 25e3b054f779..4b8ed597c59f 100644
+--- a/hotspot/src/os_cpu/linux_aarch64/vm/copy_linux_aarch64.s
++++ b/hotspot/src/os_cpu/linux_aarch64/vm/copy_linux_aarch64.s
+@@ -159,7 +159,7 @@ bwd_copy_aligned:
+         blo     bwd_copy_drain
+ 
+ bwd_copy_again:
+-        prfm    pldl1keep, [s, #-256]
++        prfum   pldl1keep, [s, #-256]
+         stp     t0, t1, [d, #-16]
+         ldp     t0, t1, [s, #-16]
+         stp     t2, t3, [d, #-32]
+diff --git a/hotspot/src/os_cpu/linux_aarch64/vm/os_linux_aarch64.cpp b/hotspot/src/os_cpu/linux_aarch64/vm/os_linux_aarch64.cpp
+index c728687a4044..e428e89baa09 100644
+--- a/hotspot/src/os_cpu/linux_aarch64/vm/os_linux_aarch64.cpp
++++ b/hotspot/src/os_cpu/linux_aarch64/vm/os_linux_aarch64.cpp
+@@ -83,19 +83,13 @@
+ #define REG_SP REG_RSP
+ #define REG_PC REG_RIP
+ #define REG_FP REG_RBP
+-#define SPELL_REG_SP "rsp"
+-#define SPELL_REG_FP "rbp"
+ #else
+ #define REG_FP 29
+ #define REG_LR 30
+-
+-#define SPELL_REG_SP "sp"
+-#define SPELL_REG_FP "x29"
+ #endif
+ 
+-address os::current_stack_pointer() {
+-  register void *esp __asm__ (SPELL_REG_SP);
+-  return (address) esp;
++NOINLINE address os::current_stack_pointer() {
++  return (address)__builtin_frame_address(0);
+ }
+ 
+ char* os::non_memory_address_word() {
+@@ -233,14 +233,8 @@ frame os::get_sender_for_C_frame(frame* fr) {
+ #endif
+ }
+ 
+-intptr_t* _get_previous_fp() {
+-  register intptr_t **ebp __asm__ (SPELL_REG_FP);
+-  return (intptr_t*) *ebp;   // we want what it points to.
+-}
+-
+-
+-frame os::current_frame() {
+-  intptr_t* fp = _get_previous_fp();
++NOINLINE frame os::current_frame() {
++  intptr_t *fp = *(intptr_t **)__builtin_frame_address(0);
+   frame myframe((intptr_t*)os::current_stack_pointer(),
+                 (intptr_t*)fp,
+                 CAST_FROM_FN_PTR(address, os::current_frame));
+@@ -259,12 +238,6 @@ frame os::current_frame() {
+ }
+ 
+ // Utility functions
+-
+-// From IA32 System Programming Guide
+-enum {
+-  trap_page_fault = 0xE
+-};
+-
+ #ifdef BUILTIN_SIM
+ extern "C" void Fetch32PFI () ;
+ extern "C" void Fetch32Resume () ;
+@@ -667,42 +640,42 @@ extern "C" {
+     return 0;
+   }
+ 
+-  void _Copy_conjoint_jshorts_atomic(jshort* from, jshort* to, size_t count) {
++  void _Copy_conjoint_jshorts_atomic(const jshort* from, jshort* to, size_t count) {
+     if (from > to) {
+-      jshort *end = from + count;
++      const jshort *end = from + count;
+       while (from < end)
+         *(to++) = *(from++);
+     }
+     else if (from < to) {
+-      jshort *end = from;
++      const jshort *end = from;
+       from += count - 1;
+       to   += count - 1;
+       while (from >= end)
+         *(to--) = *(from--);
+     }
+   }
+-  void _Copy_conjoint_jints_atomic(jint* from, jint* to, size_t count) {
++  void _Copy_conjoint_jints_atomic(const jint* from, jint* to, size_t count) {
+     if (from > to) {
+-      jint *end = from + count;
++      const jint *end = from + count;
+       while (from < end)
+         *(to++) = *(from++);
+     }
+     else if (from < to) {
+-      jint *end = from;
++      const jint *end = from;
+       from += count - 1;
+       to   += count - 1;
+       while (from >= end)
+         *(to--) = *(from--);
+     }
+   }
+-  void _Copy_conjoint_jlongs_atomic(jlong* from, jlong* to, size_t count) {
++  void _Copy_conjoint_jlongs_atomic(volatile jlong* from, volatile jlong* to, size_t count) {
+     if (from > to) {
+-      jlong *end = from + count;
++      const jlong *end = (const jlong*)from + count;
+       while (from < end)
+         os::atomic_copy64(from++, to++);
+     }
+     else if (from < to) {
+-      jlong *end = from;
++      const jlong *end = (const jlong*)from;
+       from += count - 1;
+       to   += count - 1;
+       while (from >= end)
+@@ -710,22 +683,22 @@ extern "C" {
+     }
+   }
+ 
+-  void _Copy_arrayof_conjoint_bytes(HeapWord* from,
++  void _Copy_arrayof_conjoint_bytes(const HeapWord* from,
+                                     HeapWord* to,
+                                     size_t    count) {
+     memmove(to, from, count);
+   }
+-  void _Copy_arrayof_conjoint_jshorts(HeapWord* from,
++  void _Copy_arrayof_conjoint_jshorts(const HeapWord* from,
+                                       HeapWord* to,
+                                       size_t    count) {
+     memmove(to, from, count * 2);
+   }
+-  void _Copy_arrayof_conjoint_jints(HeapWord* from,
++  void _Copy_arrayof_conjoint_jints(const HeapWord* from,
+                                     HeapWord* to,
+                                     size_t    count) {
+     memmove(to, from, count * 4);
+   }
+-  void _Copy_arrayof_conjoint_jlongs(HeapWord* from,
++  void _Copy_arrayof_conjoint_jlongs(const HeapWord* from,
+                                      HeapWord* to,
+                                      size_t    count) {
+     memmove(to, from, count * 8);
+diff --git a/hotspot/src/os/linux/vm/os_linux.inline.hpp b/hotspot/src/os/linux/vm/os_linux.inline.hpp
+index 937cd47..a6c9a87 100644
+--- a/hotspot/src/os/linux/vm/os_linux.inline.hpp
++++ b/hotspot/src/os/linux/vm/os_linux.inline.hpp
+@@ -99,18 +99,16 @@ inline int os::ftruncate(int fd, jlong length) {
+ inline struct dirent* os::readdir(DIR* dirp, dirent *dbuf)
+ {
+   dirent* p;
+-  int status;
+   assert(dirp != NULL, "just checking");
+ 
+   // NOTE: Linux readdir_r (on RH 6.2 and 7.2 at least) is NOT like the POSIX
+   // version. Here is the doc for this function:
+   // http://www.gnu.org/manual/glibc-2.2.3/html_node/libc_262.html
+ 
+-  if((status = ::readdir_r(dirp, dbuf, &p)) != 0) {
+-    errno = status;
+-    return NULL;
+-  } else
++  if((p = ::readdir(dirp)) != NULL) {
+     return p;
++  } else
++    return NULL;
+ }
+ 
+ inline int os::closedir(DIR *dirp) {

base-commit: 472706ae2f9160833951a4e4bcc4c206e03097b0
-- 
2.39.0

[bug#63069] Still relevant?

[Andreas Enge]

Hello Anthony,

as written at
   https://issues.guix.gnu.org/63327
the much simpler patch given there enables openjdk@9 to be built on my
aarch64 machine.

So can this bug be closed?

Andreas

[bug#63069] Still relevant?

[Andreas Enge]

Am Wed, Jul 26, 2023 at 04:04:37PM +0200 schrieb Andreas Enge:
>    https://issues.guix.gnu.org/63327
> the much simpler patch given there enables openjdk@9 to be built on my
> aarch64 machine.

Well, but the new phase does not apply to openjdk@10. At the very least,
it needs to be deleted there.

Andreas

[bug#63069] Still relevant?

[Andreas Enge]

Am Wed, Jul 26, 2023 at 04:23:11PM +0200 schrieb Andreas Enge:
> Well, but the new phase does not apply to openjdk@10. At the very least,
> it needs to be deleted there.

But this is not enough. Giving up on this and hoping that Chris, in cc,
has an idea...

Andreas

[bug#63069] Still relevant?

[Bruno Victal]

Hi Andreas,

On 2023-07-26 15:55, Andreas Enge wrote:
> Am Wed, Jul 26, 2023 at 04:23:11PM +0200 schrieb Andreas Enge:
>> Well, but the new phase does not apply to openjdk@10. At the very least,
>> it needs to be deleted there.
> 
> But this is not enough. Giving up on this and hoping that Chris, in cc,
> has an idea...

FWIW I think it's cleaner to use a .patch file for this rather than an extra
phase and dance of substitute-keyword-arguments/phase manipulations.


-- 
Furthermore, I consider that nonfree software must be eradicated.

Cheers,
Bruno.

[bug#63069] Alternative solution for both openjdk9 and openjdk10.

[Tim Johann]

[-- Attachment #1.1: Type: text/plain, Size: 251 bytes --]

Hi Andreas, hi Bruno, hi Chris,

See my contribution for a fix for both openjdk9 and openjdk10 using separate .patch files.  I was able to compile openjdk11 afterward applying this change.

<https://issues.guix.gnu.org/64670>

Thanks

and cheers

Tim

[-- Attachment #1.2: Type: text/html, Size: 1560 bytes --]