OpenJDK - restore version 17

Author: psumbera

OracleSolaris_OpenJDK_Builder/jdk-17.sh

@@ -0,0 +1,28 @@
+set -xe
+
+VERSION=17
+
+. common.sh
+
+BOOT_JDK="$BUILD_DIR/jdk16/build/solaris-$JDK_PLATFORM-server-release/jdk"
+PATH="/usr/bin:/usr/gnu/bin"
+
+CONFIGURE_OPTIONS+=" --enable-unlimited-crypto"
+CONFIGURE_OPTIONS+=" --enable-deprecated-ports=yes"
+CONFIGURE_OPTIONS+=" --with-boot-jdk=$BOOT_JDK"
+CONFIGURE_OPTIONS+=" --with-native-debug-symbols=none"
+CONFIGURE_OPTIONS+=" --with-toolchain-type=gcc"
+CONFIGURE_OPTIONS+=" --disable-hotspot-gtest"
+CONFIGURE_OPTIONS+=" --disable-dtrace"
+CONFIGURE_OPTIONS+=" --disable-warnings-as-errors"
+CONFIGURE_OPTIONS+=" AS=/usr/gnu/bin/as"
+CONFIGURE_OPTIONS+=" CC=$GCC"
+CONFIGURE_OPTIONS+=" CXX=$GXX"
+
+git clone ${JDK_GITHUB_REPO}/$SRC_DIR "$BUILD_DIR"/$SRC_DIR
+cd "$BUILD_DIR"/$SRC_DIR
+
+apply_patch_series
+
+PATH="$PATH" bash ./configure ${CONFIGURE_OPTIONS} 
+gmake bundles 

OracleSolaris_OpenJDK_Builder/patches-17/README-removal.txt

@@ -0,0 +1,34 @@
+15+24 was tagged after the solaris/sparc removal but didn't actually
+include it, it was based on a slightly earlier commit.
+
+However, this is a set of notes on how the removal is being addressed
+http://cr.openjdk.java.net/~mikael/webrevs/8244224/webrev.final/open/webrev/
+it's therefore relatively trivial to revert the removal at this point
+
+The patch there was:
+
+wc java-solaris-sparc.patch
+  143793  691143 5454385 java-solaris-sparc.patch
+
+And there are 953 files affected by the patch
+
+grep '^---' java-solaris-sparc.patch | awk '{print $2}' | wc -l
+     953
+
+297 of those are under test, we probably don't want them (for now, at
+any rate)
+
+214 are sparc, which again we don't want
+
+There are some mac/bsd/linux patches left after that
+
+Overall, we want less that half the files in the patch, based on
+filename alone.
+
+It turns out that gpatch couldn't cope with the original
+patch. Rebuilt it as 64-bit and it was much happier.
+
+For 15+26, I've trimmed about 2/3 of java-solaris-sparc.patch
+This is mostly about not restoring SPARC, there's still quite a lot
+more cleanup we could do. Also, we shouldn't have to separately patch
+the files we created by applying the main patch.

OracleSolaris_OpenJDK_Builder/patches-17/README.txt

@@ -0,0 +1,241 @@
+This is originally derived from the pkgsrc-joyent set for openjdk11,
+considerably modified. Cut from the jdk15 patches as of jdk15+32.
+
+See also README-zero.txt for patches to build a project zero variant.
+
+Most patches -p0
+
+JDK17:
+
+17-35
+
+RC1, no changes
+
+17-29,17-30,17-31,17-32,17-33,17-34
+
+No changes
+
+17-27,17-28
+
+No changes. Packaged 17-28.
+
+17-26
+
+First from the jdk17 stabilisation repository. Same changes as in the
+first jdk18 tag.
+
+Remove extended file io workaround, it's 32-bit only
+
+17-25
+
+SecurityManager deprecation. Added suppression annotations (the way it
+was done for Windows), and cleaned up the line number noise.
+
+17-24
+
+Line number noise.
+
+Clean up some of the compiler warnings.
+
+17-23
+
+Line number noise.
+
+os::os_exception_wrapper signature; last argument is a JavaThread not
+a Thread. No impact as we don't actually use it.
+
+17-22
+
+Signature of reserve_memory_special and pd_reserve_memory_special has
+changed to pass the page size. All we do is change the signature as
+it's a no-op on solaris.
+
+17-21
+
+Line number noise.
+
+17-20
+
+Quite a lot of changes this time around.
+
+No more aot, so remove patches for that.
+
+src/hotspot/share/cds/classListParser.cpp has moved to
+src/hotspot/share/cds/classListParser.cpp
+
+suspend/resume has been essentially gutted
+
+Assembler files are all .S not .s now; and the way it's called has
+changed, so we need to just forget about forcing the studio assembler
+
+Other changes made:
+
+Also removed solstudio, which also cleans up C_FLAG_REORDER
+
+Cleared out some of the sparc pieces
+
+Gutted ucrypto
+
+17-19
+
+.mx.jvmci has gone, we don't need it anyway
+
+Boot jdk requirement bumped to 16.
+This triggers a lot of noisy
+"SIGSEGV happened inside stack but outside yellow and red zone."
+warnings during the build, but they appear harmless.
+
+17-18
+
+Don't patch jib-profiles.js. We don't use this, there aren't any AIX
+entries, so putting ours in is a waste of effort.
+
+Remove the nbproject patches; they're either wrong or irrelevant, or both.
+
+The boolAtPut flag stuff is no more. The whole JVMFlag machinery has
+been reworked. Renamed to set_bool.
+
+Remove patching of JvmOverrideFiles.gmk when we then disable the
+patch later.
+
+17-17
+
+No significant changes
+
+17-16
+
+Some patch noise this time. Some of the SPARC-related noise simply
+removed.
+
+17-15
+
+No changes this time
+
+17-14
+
+Remove patch to c1_LIRAssembler.cpp as it's only for SPARC
+Maybe lir_pack/lir_unpack could go too?
+
+Tidied up Awt2dLibraries.gmk
+
+More centralization into os_posix: os::die() os::shutdown() os::abort()
+
+Tidy up some of the format noise
+
+17-13
+
+Remove the patches for lib-X11.m4; they're actually for S10, so we
+don't want them anyway. And remove the patch to XRBackendNative.c
+related to it which is also an S10 fixup.
+
+os::fork_and_exec has been centralized
+
+MetaspaceShared::is_in_trampoline_frame() has been removed
+
+17-12
+
+Remove saved_rbp_address_offset() from thread_solaris_x86.hpp
+
+There's also evidence of xattr support on some platforms having been
+added; it's not needed for the build so I've ignored it for now.
+
+17-11
+
+Fix memTracker.hpp; signature changed
+Don't patch deoptimization.cpp, it was SPARC-specific
+
+17-10
+
+No significant changes
+
+17-9
+
+No significant changes
+
+17-8
+
+Man os:: functions were centralized into os_posix.inline.hpp, so
+needed to remove our duplicates from os_solaris.inline.hpp and
+os_solaris.cpp. Mostly the central copy was identical, the only odd
+one was that the Solaris os::connect had some funky error handling
+
+17-7
+
+CLOCK_MONOTONIC is now just on.
+Needed to remove os::supports_monotonic_clock() from os_solaris.inline.hpp
+Remove javaTimeMillis() and javaTimeSystemUTC() and javaTimeNanos()
+and javaTimeNanos_info() from os_solaris.cpp
+Parker::park and Parker::unpark no longer compile, park.hpp has been
+modified so we need to make sure _counter is present in
+os_solaris.hpp; that makes it compile but it then crashes the jvm in there
+The posix variant removed the guarantee in PlatformParker; we need to
+do the same, fix the signature, and provide an implementation.
+(The PlatformParker piece took a lot of effort.)
+
+17-6
+
+we now have perfMemory_posix.cpp, so perfMemory_solaris.ccp should be
+removed or the build fails with multiply defined symbols
+
+17-5
+
+make/CompileJavaModules.gmk has mostly been gutted; ignore our patch
+as it seem to build fine without
+
+os.cpp needs to include alloca.h, see illumos-port-11.patch
+
+17-4
+
+src/jdk.hotspot.agent/share/classes/sun/jvm/hotspot/asm/Disassembler.java
+has been refactored, but the new version appears to do the same thing
+as we need in a platform-independent manner, so drop our patch
+
+17-3
+
+make/common/Modules.gmk has been completely redone; remove our patch
+
+17-2
+
+More memory stuff; pd_uncommit_memory pd_reserve_memory
+pd_attempt_reserve_memory_at have changed signatures, with an extra
+bool argument (that appears to be unused, at least linux and aix
+don't make any use of it; this appears to be a Mac thing)
+
+17-0/17-1
+
+Quite a lot of refactoring of the toolchain stuff.
+
+The page_size rework is really quite massive. The old _page_sizes was
+an array you manipulated directly. It's now a set. This needs various
+changes:
+
+os::Solaris::is_valid_page_size() is just _page_sizes.contains()
+
+Not strictly broken, but easy to fix: in mpss_sanity_check, we can
+simplify the getpagesizes() stuff as we know we're modern
+
+Rework listing of valid page sizes
+
+This is a bit ugly because we don't really end up using the new
+_page_sizes, but instead emulate the old array. Still, I can't find
+examples of how the new way is supposed to work on any other platform.
+
+Also need to implement print_memory_mappings, as a no-op (like AIX is)
+
+That's enough to make it compile; it blows up with an arithmetic
+exception, apparently inside apply_ergo(). In mpss_sanity_check, we
+need to make sure page_size (which is really a pointer to
+_large_page_size) is initialized to the largest valid page size.
+
+Build:
+
+env PATH=/usr/bin:/usr/sbin:/usr/sfw/bin:/usr/gnu/bin bash ./configure \
+--enable-unlimited-crypto --with-boot-jdk=/usr/jdk/instances/jdk16 \
+--with-native-debug-symbols=none \
+--with-toolchain-type=gcc \
+--disable-dtrace \
+--disable-warnings-as-errors \
+--enable-deprecated-ports=yes \
+--with-jobs=3
+
+env PATH=/usr/bin:/usr/sbin:/usr/sfw/bin:/usr/gnu/bin gmake all

OracleSolaris_OpenJDK_Builder/patches-17/Solaris11-EM_486.patch

@@ -0,0 +1,11 @@
+--- openjdk14/src/hotspot/os/solaris/os_solaris.cpp.orig	2021-08-30 09:47:00.387082887 +0000
++++ openjdk14/src/hotspot/os/solaris/os_solaris.cpp	2021-08-30 09:48:20.965475955 +0000
+@@ -1438,7 +1438,7 @@
+ 
+   static const arch_t arch_array[]={
+     {EM_386,         EM_386,     ELFCLASS32, ELFDATA2LSB, (char*)"IA 32"},
+-    {EM_486,         EM_386,     ELFCLASS32, ELFDATA2LSB, (char*)"IA 32"},
++    {EM_IAMCU,       EM_386,     ELFCLASS32, ELFDATA2LSB, (char*)"IA 32"},
+     {EM_IA_64,       EM_IA_64,   ELFCLASS64, ELFDATA2LSB, (char*)"IA 64"},
+     {EM_X86_64,      EM_X86_64,  ELFCLASS64, ELFDATA2LSB, (char*)"AMD 64"},
+     {EM_SPARC,       EM_SPARC,   ELFCLASS32, ELFDATA2MSB, (char*)"Sparc 32"},

OracleSolaris_OpenJDK_Builder/patches-17/Solaris11-TRAPNO.patch

@@ -0,0 +1,11 @@
+--- openjdk14/src/hotspot/os_cpu/solaris_x86/os_solaris_x86.cpp.orig	2021-08-30 09:50:25.617131295 +0000
++++ openjdk14/src/hotspot/os_cpu/solaris_x86/os_solaris_x86.cpp	2021-08-30 09:54:16.121332502 +0000
+@@ -618,7 +618,7 @@
+   // Furthermore, a false-positive should be harmless.
+   if (UnguardOnExecutionViolation > 0 &&
+       (sig == SIGSEGV || sig == SIGBUS) &&
+-      uc->uc_mcontext.gregs[TRAPNO] == T_PGFLT) {  // page fault
++      uc->uc_mcontext.gregs[REG32_TRAPNO] == T_PGFLT) {  // page fault
+     int page_size = os::vm_page_size();
+     address addr = (address) info->si_addr;
+     address pc = (address) uc->uc_mcontext.gregs[REG_PC];

OracleSolaris_OpenJDK_Builder/patches-17/Solaris11-memcntl.patch

@@ -0,0 +1,24 @@
+--- jdk16u-jdk-16.0.0-0/src/hotspot/os/solaris/os_solaris.cpp
++++ jdk16u-jdk-16.0.0-0/src/hotspot/os/solaris/os_solaris.cpp
+@@ -168,12 +168,6 @@
+ 
+ os::Solaris::pthread_setname_np_func_t os::Solaris::_pthread_setname_np = NULL;
+ 
+-// "default" initializers for missing libc APIs
+-extern "C" {
+-  int memcntl(caddr_t, size_t, int, caddr_t, int, int);
+-  int meminfo(const uint64_t *, int, const uint_t *, int, uint64_t *, uint_t *);
+-}
+-
+ // "default" initializers for pthread-based synchronization
+ extern "C" {
+   static int pthread_mutex_default_init(mutex_t *mx, int scope, void *arg) { memset(mx, 0, sizeof(mutex_t)); return 0; }
+@@ -2242,7 +2236,7 @@
+   mpss_struct.mha_pagesize = align;
+   mpss_struct.mha_flags = 0;
+   // Upon successful completion, memcntl() returns 0
+-  if (memcntl(start, bytes, MC_HAT_ADVISE, (caddr_t) &mpss_struct, 0, 0)) {
++  if (memcntl(start, bytes, MC_HAT_ADVISE, (void*) &mpss_struct, 0, 0)) {
+     debug_only(warning("Attempt to use MPSS failed."));
+     return false;
+   }