From gbenson at redhat.com  Thu Dec  3 07:30:57 2009
From: gbenson at redhat.com (Gary Benson)
Date: Thu, 3 Dec 2009 15:30:57 +0000
Subject: Remove dead patch
Message-ID: <20091203153057.GD3573@redhat.com>

Hi all,

The commit I just made to IcedTea6 and 7 removes a dead patch that was
no longer necessary.

Cheers,
Gary

-- 
http://gbenson.net/
-------------- next part --------------
diff -r 1d33b4733b4b -r 8c38c97e3ce8 ChangeLog
--- a/ChangeLog	Thu Dec 03 09:16:00 2009 +0000
+++ b/ChangeLog	Thu Dec 03 13:49:45 2009 +0000
@@ -1,3 +1,9 @@
+2009-12-03  Gary Benson  <gbenson at redhat.com>
+
+	* patches/icedtea-bytecodeInterpreter.patch: Removed.
+	* Makefile.am (ICEDTEA_PATCHES): Don't apply the above.
+	* HACKING: Document the above.
+
 2009-12-03  Gary Benson  <gbenson at redhat.com>
 
 	* overlays/openjdk/jdk/src/share/classes/com/sun/media/sound/:
diff -r 1d33b4733b4b -r 8c38c97e3ce8 HACKING
--- a/HACKING	Thu Dec 03 09:16:00 2009 +0000
+++ b/HACKING	Thu Dec 03 13:49:45 2009 +0000
@@ -99,8 +99,6 @@
 * icedtea-cacao.patch: For the 'java' command, create new thread depending on the current VM.
 * icedtea-cacao-no-mmap-first-page.patch: Don't mmap the first memory page.
 * cacao/no-strict-aliasing.patch: Turn off strict aliasing which causes an issue with the verifier when building with GCC 4.4 (cacao PR129).
-
-* icedtea-bytecodeInterpreter.patch: Replace fast opcodes with opc_default.
 * icedtea-ia64-bugfix.patch: Remove workaround for IA64 GCC bug.
 * icedtea-signature-iterator.patch: Add zero-specific signature handling.
 * icedtea-test-atomic-operations.patch: Add check to see if GCC's sync_lock_test works.
@@ -177,3 +175,4 @@
 * icedtea-core-zero-build.patch: Not required with HS14.
 * icedtea-bytecodeInterpreterWithChecks.patch: Same as icedtea-xslfix.patch (OpenJDK6 only, S6707485); not required with HS14.
 * icedtea-eclipse-hotspot-6614100-7b24.patch: Fix Eclipse crash (S6614100). Fixed in OpenJDK7 b29/hs13.  Not required with HS14.
+* icedtea-bytecodeInterpreter.patch: Replace fast opcodes with opc_default.  Not required with HS14.
diff -r 1d33b4733b4b -r 8c38c97e3ce8 Makefile.am
--- a/Makefile.am	Thu Dec 03 09:16:00 2009 +0000
+++ b/Makefile.am	Thu Dec 03 13:49:45 2009 +0000
@@ -516,7 +516,6 @@
 
 ICEDTEA_PATCHES = \
 	patches/icedtea-notice-safepoints.patch \
-	patches/icedtea-bytecodeInterpreter.patch \
 	patches/icedtea-signature-iterator.patch \
 	patches/icedtea-test-atomic-operations.patch \
 	patches/icedtea-zero.patch \
diff -r 1d33b4733b4b -r 8c38c97e3ce8 patches/icedtea-bytecodeInterpreter.patch
--- a/patches/icedtea-bytecodeInterpreter.patch	Thu Dec 03 09:16:00 2009 +0000
+++ /dev/null	Thu Jan 01 00:00:00 1970 +0000
@@ -1,11 +0,0 @@
-diff -Nru openjdk.orig/hotspot/src/share/vm/interpreter/bytecodeInterpreter.hpp openjdk/hotspot/src/share/vm/interpreter/bytecodeInterpreter.hpp
---- openjdk.orig/hotspot/src/share/vm/interpreter/bytecodeInterpreter.hpp	2008-10-23 08:41:04.000000000 +0100
-+++ openjdk/hotspot/src/share/vm/interpreter/bytecodeInterpreter.hpp	2008-10-30 15:48:48.000000000 +0000
-@@ -60,7 +60,6 @@
- };
- 
- class BytecodeInterpreter : StackObj {
--friend class SharedRuntime;
- friend class AbstractInterpreterGenerator;
- friend class CppInterpreterGenerator;
- friend class InterpreterGenerator;

From gbenson at redhat.com  Thu Dec  3 08:06:00 2009
From: gbenson at redhat.com (Gary Benson)
Date: Thu, 3 Dec 2009 16:06:00 +0000
Subject: Remove ancient workaround
Message-ID: <20091203160600.GA9977@redhat.com>

Hi all,

This commit removes a workaround for building against versions
of Zero from April 2008 and earlier (or against the PowerPC
interpreter -- remember that?)  Please don't anybody tell me
they have a build that old they need to support...

Cheers,
Gary

-- 
http://gbenson.net/
-------------- next part --------------
diff -r 8c38c97e3ce8 -r bc6d4b57cacf ChangeLog
--- a/ChangeLog	Thu Dec 03 13:49:45 2009 +0000
+++ b/ChangeLog	Thu Dec 03 15:58:48 2009 +0000
@@ -1,3 +1,7 @@
+2009-12-03  Gary Benson  <gbenson at redhat.com>
+
+	* patches/icedtea-core-build.patch: Remove a very old workaround.
+
 2009-12-03  Gary Benson  <gbenson at redhat.com>
 
 	* patches/icedtea-bytecodeInterpreter.patch: Removed.
diff -r 8c38c97e3ce8 -r bc6d4b57cacf patches/icedtea-core-build.patch
--- a/patches/icedtea-core-build.patch	Thu Dec 03 13:49:45 2009 +0000
+++ b/patches/icedtea-core-build.patch	Thu Dec 03 15:58:48 2009 +0000
@@ -44,38 +44,3 @@
  
  #  NOT Windows ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ NOT Windows
  
-diff -ru openjdk.orig/jdk/make/common/Defs-linux.gmk openjdk/jdk/make/common/Defs-linux.gmk
---- openjdk.orig/jdk/make/common/Defs-linux.gmk	2007-11-08 11:35:35.000000000 +0000
-+++ openjdk/jdk/make/common/Defs-linux.gmk	2007-11-08 11:36:15.000000000 +0000
-@@ -258,7 +258,12 @@ endif
- # -L to the -ljvm, this is because -ljava depends on -ljvm, whereas
- # the library itself should not.
- #
--VM_NAME         = server
-+VM_NAME = $(shell for dir in server client core; do \
-+                    if [ -d $(LIBDIR)/$(LIBARCH)/$$dir ]; then \
-+		      echo $$dir; \
-+		      break; \
-+		    fi; \
-+		  done)
- JVMLIB          = -L$(LIBDIR)/$(LIBARCH)/$(VM_NAME) -ljvm
- JAVALIB         = -ljava $(JVMLIB)
- 
-diff -ru openjdk.orig/corba/make/common/Defs-linux.gmk openjdk/corba/make/common/Defs-linux.gmk
---- openjdk.orig/corba/make/common/Defs-linux.gmk	Tue Feb 05 15:52:33 2008 +0000
-+++ openjdk/corba/make/common/Defs-linux.gmk	Mon Feb 11 10:01:46 2008 +0000
-@@ -250,7 +254,12 @@ endif
- # -L to the -ljvm, this is because -ljava depends on -ljvm, whereas
- # the library itself should not.
- #
--VM_NAME         = server
-+VM_NAME = $(shell for dir in server client core; do \
-+                    if [ -d $(BOOTDIR)/jre/lib/$(LIBARCH)/$$dir ]; then \
-+		      echo $$dir; \
-+		      break; \
-+		    fi; \
-+		  done)
- JVMLIB          = -L$(BOOTDIR)/jre/lib/$(LIBARCH)/$(VM_NAME) -ljvm
- JAVALIB         = -L$(BOOTDIR)/jre/lib/$(LIBARCH) -ljava $(JVMLIB)
- 
-

From gbenson at redhat.com  Thu Dec  3 08:41:46 2009
From: gbenson at redhat.com (Gary Benson)
Date: Thu, 3 Dec 2009 16:41:46 +0000
Subject: Remove more dead patches
Message-ID: <20091203164146.GB9977@redhat.com>

Hi all,

This commit removes the ancient workaround I mentioned before from
IcedTea7, and another dead patch that is not required to build Zero.

Cheers,
Gary

-- 
http://gbenson.net/
-------------- next part --------------
diff -r 21376f9eed89 -r 2deccfc8b64f ChangeLog
--- a/ChangeLog	Thu Dec 03 13:53:28 2009 +0000
+++ b/ChangeLog	Thu Dec 03 16:39:28 2009 +0000
@@ -1,3 +1,10 @@
+2009-12-03  Gary Benson  <gbenson at redhat.com>
+
+	* patches/icedtea-core-build.patch: Removed.
+	* patches/icedtea-zero-build.patch: Likewise.
+	* Makefile.am (ICEDTEA_PATCHES): Don't apply the above.
+	* HACKING: Document the above.
+
 2009-12-03  Gary Benson  <gbenson at redhat.com>
 
 	* patches/hotspot/icedtea-bytecodeInterpreter.patch: Removed.
diff -r 21376f9eed89 -r 2deccfc8b64f HACKING
--- a/HACKING	Thu Dec 03 13:53:28 2009 +0000
+++ b/HACKING	Thu Dec 03 16:39:28 2009 +0000
@@ -15,7 +15,6 @@
   Partially fixed by addition of discardMark() in OpenJDK7.
 * icedtea-copy-plugs.patch: Add IcedTea's 'plugs' for Java WebStart support, add
   cast to getAnnotation and trapAuthenticationFailure variable to com.sun.jmx.snmp.SnmpPduTrap.
-* icedtea-core-build.patch: Add support for building IcedTea core VM.
 * icedtea-debuginfo.patch: Add -g option to build to generate debugging information.
 * icedtea-demos.patch: Fix building of JVMTI demo.
 * icedtea-float-double-trailing-zeros.patch: Remove trailing zeros from Double/Float (PR29/30)
@@ -49,7 +48,6 @@
 * icedtea-use-system-tzdata.patch: Use timezone data from the system (PR70/S6593486).
 * icedtea-version.patch: Replace name and bug URL with IcedTea versions.
 * icedtea-webservices.patch: Add applet support.
-* icedtea-zero-build.patch: Add support for the zero assembler build.
 * icedtea-shark-build.patch: Add support for building the Shark JIT.
 * icedtea-alpha-fixes.patch: Fix build issues on alpha-linux.
 * icedtea-arch.patch: Add support for additional architectures.
@@ -198,3 +196,5 @@
 * icedtea-bytecodeInterpreterWithChecks.patch: Same as icedtea-xslfix.patch (OpenJDK6 only, S6707485).
 * icedtea-eclipse-hotspot-6614100-7b24.patch: Fix Eclipse crash (S6614100). Fixed in OpenJDK7 b29/hs13.
 * icedtea-bytecodeInterpreter.patch: Replace fast opcodes with opc_default.
+* icedtea-core-build.patch: Add support for building IcedTea core VM.
+* icedtea-zero-build.patch: Add support for the zero assembler build.
diff -r 21376f9eed89 -r 2deccfc8b64f Makefile.am
--- a/Makefile.am	Thu Dec 03 13:53:28 2009 +0000
+++ b/Makefile.am	Thu Dec 03 16:39:28 2009 +0000
@@ -254,8 +254,6 @@
 	patches/hotspot/$(HSBUILD)/icedtea-memory-limits.patch \
 	patches/icedtea-sunsrc.patch \
 	patches/icedtea-libraries.patch \
-	patches/icedtea-core-build.patch \
-	patches/icedtea-zero-build.patch \
 	patches/icedtea-print-lsb-release.patch \
 	patches/icedtea-jpegclasses.patch \
 	patches/icedtea-uname.patch \
diff -r 21376f9eed89 -r 2deccfc8b64f patches/icedtea-core-build.patch
--- a/patches/icedtea-core-build.patch	Thu Dec 03 13:53:28 2009 +0000
+++ /dev/null	Thu Jan 01 00:00:00 1970 +0000
@@ -1,34 +0,0 @@
-diff -Nru openjdk.orig/corba/make/common/Defs-linux.gmk openjdk/corba/make/common/Defs-linux.gmk
---- openjdk.orig/corba/make/common/Defs-linux.gmk	2009-07-15 17:35:01.000000000 +0100
-+++ openjdk/corba/make/common/Defs-linux.gmk	2009-07-23 17:41:01.000000000 +0100
-@@ -258,7 +258,12 @@
- # -L to the -ljvm, this is because -ljava depends on -ljvm, whereas
- # the library itself should not.
- #
--VM_NAME         = server
-+VM_NAME = $(shell for dir in server client core; do \
-+                    if [ -d $(BOOTDIR)/jre/lib/$(LIBARCH)/$$dir ]; then \
-+		      echo $$dir; \
-+		      break; \
-+		    fi; \
-+		  done)
- JVMLIB          = -L$(BOOTDIR)/jre/lib/$(LIBARCH)/$(VM_NAME) -ljvm
- JAVALIB         = -L$(BOOTDIR)/jre/lib/$(LIBARCH) -ljava $(JVMLIB)
- 
-diff -Nru openjdk.orig/jdk/make/common/Defs-linux.gmk openjdk/jdk/make/common/Defs-linux.gmk
---- openjdk.orig/jdk/make/common/Defs-linux.gmk	2009-07-15 17:35:01.000000000 +0100
-+++ openjdk/jdk/make/common/Defs-linux.gmk	2009-07-23 17:41:01.000000000 +0100
-@@ -282,7 +282,12 @@
- # -L to the -ljvm, this is because -ljava depends on -ljvm, whereas
- # the library itself should not.
- #
--VM_NAME         = server
-+VM_NAME = $(shell for dir in server client core; do \
-+                    if [ -d $(LIBDIR)/$(LIBARCH)/$$dir ]; then \
-+		      echo $$dir; \
-+		      break; \
-+		    fi; \
-+		  done)
- JVMLIB          = -L$(LIBDIR)/$(LIBARCH)/$(VM_NAME) -ljvm
- JAVALIB         = -ljava $(JVMLIB)
- 
diff -r 21376f9eed89 -r 2deccfc8b64f patches/icedtea-zero-build.patch
--- a/patches/icedtea-zero-build.patch	Thu Dec 03 13:53:28 2009 +0000
+++ /dev/null	Thu Jan 01 00:00:00 1970 +0000
@@ -1,15 +0,0 @@
-diff -Nru openjdk.orig/jdk/make/common/shared/Compiler-gcc.gmk openjdk/jdk/make/common/shared/Compiler-gcc.gmk
---- openjdk.orig/jdk/make/common/shared/Compiler-gcc.gmk	2009-04-14 18:31:09.000000000 +0100
-+++ openjdk/jdk/make/common/shared/Compiler-gcc.gmk	2009-04-14 20:32:47.000000000 +0100
-@@ -66,6 +66,11 @@
-   else
-     CXX            = $(COMPILER_PATH)g++$(GCC_SUFFIX)
-   endif
-+  ifdef ZERO_BUILD
-+    # zero
-+    REQUIRED_CC_VER = 3.2
-+    REQUIRED_GCC_VER = 3.2.*
-+  endif
-   # Option used to create a shared library
-   SHARED_LIBRARY_FLAG = -shared -mimpure-text
-   SUN_COMP_VER := $(shell $(CC) --verbose 2>&1 )

From gbenson at redhat.com  Fri Dec  4 10:33:03 2009
From: gbenson at redhat.com (Gary Benson)
Date: Fri, 4 Dec 2009 18:33:03 +0000
Subject: Zero patch rearrangement in icedtea6
Message-ID: <20091204183302.GB17799@redhat.com>

Hi all,

At some point Zero is going to end up in a HSX tarball, and no doubt
shortly afterwards in icedtea6.  I've been working on a refactoring
that will make the transition easier to manage.

The attached patch replaces the various individual Zero patches with
larger patches numbered with the Sun bug id that they were committed
to OpenJDK7 with.  This means that, once that bug's code ends up in an
HSX tarball, we can simply remove one patch to fix IcedTea.  It also
means that, if OpenJDK 6 is about to rebase on a new HotSpot, we have
patches ready made for the non-HotSpot Zero stuff that needs to be in
that tarball.

As well as the above, the attached patch upgrades icedtea6's Zero from
the original, IcedTea way of building Zero to the new, much nicer
OpenJDK7 way of doing it.  This means that dependencies between header
files are better tracked, thanks to the new includeDB, and also that
there are now no differences between icedtea6's Zero and OpenJDK7's
except for Ed's interpreter and the odd little bit here and there
that's required for the newer HotSpot.

I'm currently testing this with a bunch of different configurations
(amd64/zero, ppc/shark, ppc64/hs16/zero and s390x/zero).  Could people
please test this patch on the configurations they care about, to make
sure I didn't break anything.

Thanks (and have a good weekend!)
Gary

-- 
http://gbenson.net/
-------------- next part --------------
A non-text attachment was scrubbed...
Name: icedtea6-zero-upgrade.patch.gz
Type: application/x-gzip
Size: 24552 bytes
Desc: not available
Url : http://mail.openjdk.java.net/pipermail/zero-dev/attachments/20091204/848a3eec/attachment-0001.bin 

From gbenson at redhat.com  Tue Dec  8 03:19:27 2009
From: gbenson at redhat.com (Gary Benson)
Date: Tue, 8 Dec 2009 11:19:27 +0000
Subject: Zero patch consolidation
Message-ID: <20091208111927.GB3422@redhat.com>

Hi all,

So the Zero patch consolidation I've been working on is now complete.

As of now, the only patches relating to Zero in icedtea6 are in the
patches/zero/ directory.  Each one is numbered with the Sun bug ID
with which it was pushed to OpenJDK7, and each one is generated from
the OpenJDK7 forest into which it was initially pushed.

The rationale behind this is that at some point in the future
changesets relating to Zero may appear in OpenJDK6 or HotSpot Express
tarballs.  When this happens, patches will fail to apply.  By grouping
the patches in this way, only one patch needs removing for each
changeset.  At some further point in the future Zero itself may appear
in an OpenJDK6 or HotSpot Express tarball; when this happens the
relevant files under ports/ will be removed (the ARM interpreter will
need turning into a patch) and changes to Zero itself will be handled
in this way.

Please don't use patches/zero for other files.

I haven't mirrored this change in IcedTea7; that has very few Zero
patches left, and after b78 (or maybe b77) they will all be gone,
so it's not like it's something I'll have to keep track of long term
like I do for IcedTea6.

Cheers,
Gary

-- 
http://gbenson.net/

From gnu_andrew at member.fsf.org  Tue Dec  8 05:42:43 2009
From: gnu_andrew at member.fsf.org (Andrew John Hughes)
Date: Tue, 8 Dec 2009 13:42:43 +0000
Subject: Zero patch consolidation
In-Reply-To: <20091208111927.GB3422@redhat.com>
References: <20091208111927.GB3422@redhat.com>
Message-ID: <17c6771e0912080542x134f3b2eoea8d284ae2a2cada@mail.gmail.com>

2009/12/8 Gary Benson <gbenson at redhat.com>:
> Hi all,
>
> So the Zero patch consolidation I've been working on is now complete.
>
> As of now, the only patches relating to Zero in icedtea6 are in the
> patches/zero/ directory. ?Each one is numbered with the Sun bug ID
> with which it was pushed to OpenJDK7, and each one is generated from
> the OpenJDK7 forest into which it was initially pushed.
>
> The rationale behind this is that at some point in the future
> changesets relating to Zero may appear in OpenJDK6 or HotSpot Express
> tarballs. ?When this happens, patches will fail to apply. ?By grouping
> the patches in this way, only one patch needs removing for each
> changeset. ?At some further point in the future Zero itself may appear
> in an OpenJDK6 or HotSpot Express tarball; when this happens the
> relevant files under ports/ will be removed (the ARM interpreter will
> need turning into a patch) and changes to Zero itself will be handled
> in this way.
>
> Please don't use patches/zero for other files.
>

I like this.  Anything that makes working out which patches have been
merged and removing them easier gets a big smile from me.  I'm also
strongly in favour of organising the patches directory a bit better,
especially if it makes it clearer which stuff is going to end up in
OpenJDK6 eventually and which stuff we'll be maintaining for life.

> I haven't mirrored this change in IcedTea7; that has very few Zero
> patches left, and after b78 (or maybe b77) they will all be gone,
> so it's not like it's something I'll have to keep track of long term
> like I do for IcedTea6.
>

b77 is already out, so I just need to get round to updating IcedTea7
to build against it, dropping the ia64 patch and part of the uname
patch (the rest I'll stick in a Debian subdir as it's specific to that
platform).
I hope b78 will see a HotSpot promotion and IcedTea7 will be clear of
zero patches for the new year.

> Cheers,
> Gary
>
> --
> http://gbenson.net/
>



-- 
Andrew :-)

Free Java Software Engineer
Red Hat, Inc. (http://www.redhat.com)

Support Free Java!
Contribute to GNU Classpath and the OpenJDK
http://www.gnu.org/software/classpath
http://openjdk.java.net

PGP Key: 94EFD9D8 (http://subkeys.pgp.net)
Fingerprint: F8EF F1EA 401E 2E60 15FA  7927 142C 2591 94EF D9D8

From xerxes at zafena.se  Thu Dec 10 06:00:33 2009
From: xerxes at zafena.se (=?ISO-8859-1?Q?Xerxes_R=E5nby?=)
Date: Thu, 10 Dec 2009 15:00:33 +0100
Subject: Native method wrapper generation
In-Reply-To: <20091111163839.GL3256@redhat.com>
References: <20091111163839.GL3256@redhat.com>
Message-ID: <4B20FF01.5060108@zafena.se>

Gary Benson wrote:
> Hi all,
>
> This commit adds native wrapper generation to Shark.  It also removes
> support for LLVM < 2.6, because the new code requires it.  I'll write
> this up in more detail tomorrow with some benchmarks.
>
> Cheers,
> Gary
>
>   
Hi

+void SharkCompiler::generate_native_code(SharkEntry* entry,
...
+#if SHARK_LLVM_VERSION >= 27
+      if (!fnmatch(SharkPrintAsmOf, name, 0)) {
+        llvm::SetCurrentDebugType(X86_ONLY("x86-emitter") NOT_X86("jit"));
+        llvm::DebugFlag = true;
+      }
+      else {
+        llvm::SetCurrentDebugType(""); 
+        llvm::DebugFlag = false;
+      }
+#else

This part of the code dont work with llvm 2.7svn *release* builds since

llvm::DebugFlag

and

llvm::SetCurrentDebugType

are only defined in the llvm headers for llvm debug builds.

This was only a minor issue before the shark rewrite since the code then
where only included in non product hotspot builds, now it fail on
product builds as well in combination with a llvm 2.7svn release build.
One way to fix this would be to introduce a new debug flag based on the
info from
llvm-config --build-mode
that are set to true when llvm-config --build-mode spits out "Debug" ,
and then conditionalise the code block to be compiled if the new debug
flag are percent.

Cheers
Xerxes


From aph at redhat.com  Thu Dec 10 06:20:05 2009
From: aph at redhat.com (Andrew Haley)
Date: Thu, 10 Dec 2009 14:20:05 +0000
Subject: Native method wrapper generation
In-Reply-To: <4B20FF01.5060108@zafena.se>
References: <20091111163839.GL3256@redhat.com> <4B20FF01.5060108@zafena.se>
Message-ID: <4B210395.5050804@redhat.com>

Xerxes R?nby wrote:
> Gary Benson wrote:
>> Hi all,
>>
>> This commit adds native wrapper generation to Shark.  It also removes
>> support for LLVM < 2.6, because the new code requires it.  I'll write
>> this up in more detail tomorrow with some benchmarks.
> 
> +void SharkCompiler::generate_native_code(SharkEntry* entry,
> ...
> +#if SHARK_LLVM_VERSION >= 27
> +      if (!fnmatch(SharkPrintAsmOf, name, 0)) {
> +        llvm::SetCurrentDebugType(X86_ONLY("x86-emitter") NOT_X86("jit"));
> +        llvm::DebugFlag = true;
> +      }
> +      else {
> +        llvm::SetCurrentDebugType(""); 
> +        llvm::DebugFlag = false;
> +      }
> +#else
> 
> This part of the code dont work with llvm 2.7svn *release* builds since
> 
> llvm::DebugFlag
> 
> and
> 
> llvm::SetCurrentDebugType
> 
> are only defined in the llvm headers for llvm debug builds.
> 
> This was only a minor issue before the shark rewrite since the code then
> where only included in non product hotspot builds, now it fail on
> product builds as well in combination with a llvm 2.7svn release build.
> One way to fix this would be to introduce a new debug flag based on the
> info from
> llvm-config --build-mode
> that are set to true when llvm-config --build-mode spits out "Debug" ,
> and then conditionalise the code block to be compiled if the new debug
> flag are percent.

But the debug dumps aren't just for people who build their own
OpenJDK, they're useful for debugging the installed OpenJDK and for
people who are just writing apps and want to look at the generated
code.  This is a bad decision on the part of LLVM, but I guess we can
live with it as long as the LLVM packages we use are build with
debugging.

Andrew.

From gbenson at redhat.com  Tue Dec 22 02:49:46 2009
From: gbenson at redhat.com (Gary Benson)
Date: Tue, 22 Dec 2009 10:49:46 +0000
Subject: Command line options tester
Message-ID: <20091222104946.GB3312@redhat.com>

Hi all,

This commit adds a command line options tester I've been working on.
It won't run on IcedTea or OpenJDK as is; it requires a patch which I
mailed to hotspot-dev that doesn't look like anyone's going to look at
before next year.  But I'm committing this now so I don't lose it :)

Cheers,
Gary

-- 
http://gbenson.net/
-------------- next part --------------
diff -r 40a336c2526b -r d48415f1ab6f ChangeLog
--- a/ChangeLog	Sat Dec 19 13:55:02 2009 +0100
+++ b/ChangeLog	Tue Dec 22 11:42:03 2009 +0100
@@ -1,3 +1,8 @@
+2009-12-22  Gary Benson  <gbenson at redhat.com>
+
+	* contrib/checkopts/checkopts: New file.
+	* contrib/checkopts/Test.java: Likewise.
+
 2009-12-19  Matthias Klose  <doko at ubuntu.com>
 
 	* patches/icedtea-parisc.patch: Fix sanity check for parisc.
diff -r 40a336c2526b -r d48415f1ab6f contrib/checkopts/Test.java
--- /dev/null	Thu Jan 01 00:00:00 1970 +0000
+++ b/contrib/checkopts/Test.java	Tue Dec 22 11:42:03 2009 +0100
@@ -0,0 +1,5 @@
+class Test {
+  public static void main(String[] args) {
+    System.out.println("Hello world");
+  }
+}
diff -r 40a336c2526b -r d48415f1ab6f contrib/checkopts/checkopts
--- /dev/null	Thu Jan 01 00:00:00 1970 +0000
+++ b/contrib/checkopts/checkopts	Tue Dec 22 11:42:03 2009 +0100
@@ -0,0 +1,146 @@
+#!/usr/bin/env python
+
+import os
+import subprocess
+import sys
+
+class Flag:
+    def __init__(self, line):
+        line = line.split()
+        self.type = line.pop(0)
+        self.name = line.pop(0)
+        assert line[0] in ("=", ":=")
+        line = " ".join(line[1:])
+        self.is_diagnostic = False
+        if line.endswith("}"):
+            index = line.rfind("{")
+            if index != -1:
+                self.is_diagnostic = "diagnostic" in line[index + 1:-1].split()
+                line = line[:index]
+        self.value = line.strip()
+
+    @property
+    def test_values(self):
+        if self.type == "bool":
+            if self.name in (
+                "ExtendedDTraceProbes", # Solaris only
+                "RequireSharedSpaces",  # Not set up
+                "PauseAtStartup"):      # Just don't...
+                return (False,)
+            return (False, True)
+        return ()
+        
+    def option(self, value):
+        if self.type == "bool":
+            return "-XX:%s%s" % (value and "+" or "-", self.name)
+        elif self.type.endswith("intx"):
+            return "-XX:%s=%d" % (self.name, value)
+        raise ValueError, self.type
+
+class TestFailure(Exception):
+    pass
+
+class Main:
+    def __init__(self, java):
+        self.java = os.path.realpath(java)
+        self.base = os.path.dirname(os.path.realpath(sys.argv[0]))
+        if not os.path.exists(os.path.join(self.base, "Test.class")):
+            out, err = subprocess.Popen(
+                [self.java + "c", "Test.java"],
+                stdout = subprocess.PIPE,
+                stderr = subprocess.PIPE,
+                cwd    = self.base).communicate()
+            if out or err:
+                sys.stdout.write(out)
+                sys.stderr.write(err)
+                sys.exit(1)
+        self.passes = self.fails = 0
+        for flag in self.read_flags():
+            for value in flag.test_values:
+                self.test(flag, value)
+        print "%d passes, %d fails" % (self.passes, self.fails)
+
+    def read_flags(self):
+        out, err = subprocess.Popen(
+            [self.java, "-XX:+UnlockDiagnosticVMOptions", "-Xprintflags"],
+            stdout = subprocess.PIPE,
+            stderr = subprocess.PIPE).communicate()
+        if err:
+            sys.stderr.write(err)
+            sys.exit(1)
+        lines = out.rstrip().split("\n")
+        line = lines.pop(0)
+        if line != "[Global flags]":
+            print >>sys.stderr, "error: unexpected output %s" % repr(line)
+            sys.exit(1)
+        flags = {}
+        for line in lines:
+            flag = Flag(line)
+            assert not flags.has_key(flag.name)
+            flags[flag.name] = flag
+        flags = flags.items()
+        flags.sort()
+        return [flag for name, flag in flags]
+
+    def test(self, flag, value):
+        for dir in xrange(1000):
+            dir = os.path.join(self.base, "results", flag.name, "%03d" % dir)
+            if not os.path.isdir(dir):
+                break
+        os.makedirs(dir)
+        cmd = [self.java]
+        if flag.is_diagnostic:
+            cmd.append("-XX:+UnlockDiagnosticVMOptions")
+        if flag.name.startswith("CMS"):
+            cmd.append("-XX:+UseConcMarkSweepGC")
+        flag = flag.option(value)
+        cmd.append(flag)
+        cmd.extend(("-cp", self.base, "Test"))
+        print "%-56s %s" % (flag, dir[len(self.base + os.sep):])
+        print >>open(os.path.join(dir, "cmd"), "w"), " ".join(cmd)
+        out, err = subprocess.Popen(
+            cmd,
+            stdout = subprocess.PIPE,
+            stderr = subprocess.PIPE,
+            cwd    = dir).communicate()
+        open(os.path.join(dir, "out"), "w").write(out)
+        open(os.path.join(dir, "err"), "w").write(err)
+        try:
+            self.check(dir, out, err)
+            self.passes += 1
+        except TestFailure, failure:
+            print "\x1B[1;31m  FAIL: %s\x1B[0m" % failure
+            self.fails += 1
+
+    def fail(self, msg):
+        raise TestFailure(msg)
+
+    def check(self, dir, out, err):
+        name = os.path.basename(os.path.dirname(dir))
+        expect = ["cmd", "err", "out"]
+        actual = os.listdir(dir)
+        if name in ("LogCompilation", "LogVMOutput"):
+            if "hotspot.log" in actual:
+                expect.append("hotspot.log")
+        if name == "PerfDataSaveToFile":
+            for item in actual:
+                if item.startswith("hsperfdata_"):
+                    expect.append(item)
+                    break
+        expect.sort()
+        actual.sort()
+        while expect != actual:
+            for item in actual:
+                if item not in expect:
+                    self.fail("found %s" % item)
+            self.fail("directory contents not as expected")
+        if err:
+            self.fail("err not as expected")
+        if not "Hello world" in out.split("\n"):
+            self.fail("out not as expected")
+
+if __name__ == "__main__":
+    if len(sys.argv) != 2:
+        print >>sys.stderr, "usage: %s /path/to/jdk/bin/java" % sys.argv[0]
+        sys.exit(1)
+    Main(sys.argv[1])