Mono

Mono

[Hodgson, Simon]

I've built Mono for OpenEmbedded to run on an arm processor. But when I try to run just a simple Hello World test I get: 

mini-codegen.c:1073:mono_local_regalloc: assertion failed: (ins->dreg != -1) 

Does anyone have any suggestion as to what I might be doing wrong? 

mono -V produces:
root@at91sam9g20ek:~# mono -V
Mono JIT compiler version 2.4.2.3 (tarball Tue Nov  3 03:08:58 EST 2009)
Copyright (C) 2002-2008 Novell, Inc and Contributors. www.mono-project.com
        TLS:           normal
        GC:            Included Boehm (with typed GC)
        SIGSEGV:       normal
        Notifications: epoll
        Architecture:  armel,vfp
        Disabled:      none

Is there anymore info perhaps I can provide?

Thanks 


Simon

Re: Mono

[Koen Kooi]

On 04-11-09 21:11, Hodgson, Simon wrote:
> I've built Mono for OpenEmbedded to run on an arm processor. But when I try to run just a simple Hello World test I get:
>
> mini-codegen.c:1073:mono_local_regalloc: assertion failed: (ins->dreg != -1)
>
> Does anyone have any suggestion as to what I might be doing wrong?

I'm building locally with:

koen@dominion:/OE/org.openembedded.dev/recipes/asio$ git diff ../mono
diff --git a/recipes/mono/mono_2.4.2.3.bb b/recipes/mono/mono_2.4.2.3.bb
index 5e659f3..261d090 100644
--- a/recipes/mono/mono_2.4.2.3.bb
+++ b/recipes/mono/mono_2.4.2.3.bb
@@ -2,7 +2,7 @@ require mono-${PV}.inc

  DEPENDS = "mono-native mono-mcs-intermediate glib-2.0 perl-native"

-PR = "${INC_PR}.1"
+PR = "${INC_PR}.2"

  # mono makes use of non-thumb-compatible inline asm.
  ARM_INSTRUCTION_SET = "arm"
@@ -10,7 +10,7 @@ ARM_INSTRUCTION_SET = "arm"
  SRC_URI += "file://configure.patch;patch=1\
             file://mini-arm.patch;patch=1\
              file://0001-remove-docs-dir-from-build.patch;patch=1 \
-            file://mono-cross-compilation-for-amd64.patch;patch=1 \
+#            file://mono-cross-compilation-for-amd64.patch;patch=1 \
             "

  # Per http://www.mono-project.com/Mono:ARM

That seems to fix most issues for me. I suspect it breaks for amd64 
targets, but I don't care about those (yet).

regards,

Koen

Re: Mono

[Hodgson, Simon]

________________________________________
From: openembedded-devel-bounces@lists.openembedded.org [openembedded-devel-bounces@lists.openembedded.org] On Behalf Of Koen Kooi [k.kooi@student.utwente.nl]
Sent: 04 November 2009 20:36
To: openembedded-devel@lists.openembedded.org
Subject: Re: [oe] Mono

On 04-11-09 21:11, Hodgson, Simon wrote:
> I've built Mono for OpenEmbedded to run on an arm processor. But when I try to run just a simple Hello World test I get:
>
> mini-codegen.c:1073:mono_local_regalloc: assertion failed: (ins->dreg != -1)
>
> Does anyone have any suggestion as to what I might be doing wrong?

I'm building locally with:

koen@dominion:/OE/org.openembedded.dev/recipes/asio$ git diff ../mono
diff --git a/recipes/mono/mono_2.4.2.3.bb b/recipes/mono/mono_2.4.2.3.bb
index 5e659f3..261d090 100644
--- a/recipes/mono/mono_2.4.2.3.bb
+++ b/recipes/mono/mono_2.4.2.3.bb
@@ -2,7 +2,7 @@ require mono-${PV}.inc

  DEPENDS = "mono-native mono-mcs-intermediate glib-2.0 perl-native"

-PR = "${INC_PR}.1"
+PR = "${INC_PR}.2"

  # mono makes use of non-thumb-compatible inline asm.
  ARM_INSTRUCTION_SET = "arm"
@@ -10,7 +10,7 @@ ARM_INSTRUCTION_SET = "arm"
  SRC_URI += "file://configure.patch;patch=1\
             file://mini-arm.patch;patch=1\
              file://0001-remove-docs-dir-from-build.patch;patch=1 \
-            file://mono-cross-compilation-for-amd64.patch;patch=1 \
+#            file://mono-cross-compilation-for-amd64.patch;patch=1 \
             "

  # Per http://www.mono-project.com/Mono:ARM

That seems to fix most issues for me. I suspect it breaks for amd64
targets, but I don't care about those (yet).

regards,

Koen


That fixed it - thanks!

Simon

Re: Mono

[Phil Blundell]

> -            file://mono-cross-compilation-for-amd64.patch;patch=1 \
> +#            file://mono-cross-compilation-for-amd64.patch;patch=1 \

That patch (mono-cross-compilation-for-amd64) does look pretty bogus.
The file header says:

        "This is a workaround for bug 515050 
        https://bugzilla.novell.com/show_bug.cgi?id=515050#c2 
         
        Real fix is something else which did not make into 2.4.2 
        http://lists.ximian.com/pipermail/mono-patches/2009-June/151386.html 
        http://lists.ximian.com/pipermail/mono-patches/2009-June/151387.html 
         
        but we are ok with workaround its safe."
        
... but it's hard to see how forcing MONO_ARCH_SUPPORT_SIMD_INTRINSICS
unconditionally really counts as "safe": this seems almost certain to
break on any target that doesn't actually have that support.  If this
patch were moved to SRC_URI_append_amd64 then it'd probably be ok, but
the patches from those other two URLs look like a better solution.

p.

Re: Mono

[Khem Raj]

On Wed, Nov 4, 2009 at 2:02 PM, Phil Blundell <philb@gnu.org> wrote:
>> -            file://mono-cross-compilation-for-amd64.patch;patch=1 \
>> +#            file://mono-cross-compilation-for-amd64.patch;patch=1 \
>
> That patch (mono-cross-compilation-for-amd64) does look pretty bogus.
> The file header says:
>
>        "This is a workaround for bug 515050
>        https://bugzilla.novell.com/show_bug.cgi?id=515050#c2
>
>        Real fix is something else which did not make into 2.4.2
>        http://lists.ximian.com/pipermail/mono-patches/2009-June/151386.html
>        http://lists.ximian.com/pipermail/mono-patches/2009-June/151387.html
>
>        but we are ok with workaround its safe."

that was the remarks from the threads which discussed the above bug. I
forgot to include the complete reference.
I was merely going by the comments in the thread.
cbrake had a locally disable this patch and tested but did not yet commit it.

>
> ... but it's hard to see how forcing MONO_ARCH_SUPPORT_SIMD_INTRINSICS
> unconditionally really counts as "safe": this seems almost certain to
> break on any target that doesn't actually have that support.  If this
> patch were moved to SRC_URI_append_amd64 then it'd probably be ok, but
> the patches from those other two URLs look like a better solution.
>
> p.
>
>
>
> _______________________________________________
> Openembedded-devel mailing list
> Openembedded-devel@lists.openembedded.org
> http://lists.linuxtogo.org/cgi-bin/mailman/listinfo/openembedded-devel
>

Re: Mono

[Khem Raj]

[-- Attachment #1: Type: text/plain, Size: 988 bytes --]

On Wed, Nov 4, 2009 at 12:11 PM, Hodgson, Simon
<Simon.Hodgson@nortechonline.co.uk> wrote:
> I've built Mono for OpenEmbedded to run on an arm processor. But when I try to run just a simple Hello World test I get:
>
> mini-codegen.c:1073:mono_local_regalloc: assertion failed: (ins->dreg != -1)
>
> Does anyone have any suggestion as to what I might be doing wrong?
>
> mono -V produces:
> root@at91sam9g20ek:~# mono -V
> Mono JIT compiler version 2.4.2.3 (tarball Tue Nov  3 03:08:58 EST 2009)
> Copyright (C) 2002-2008 Novell, Inc and Contributors. www.mono-project.com
>        TLS:           normal
>        GC:            Included Boehm (with typed GC)
>        SIGSEGV:       normal
>        Notifications: epoll
>        Architecture:  armel,vfp
>        Disabled:      none
>
> Is there anymore info perhaps I can provide?

Can you apply the attached patch then rebuild mono and retry your test.

Thanks

-Khem

[-- Attachment #2: patch --]
[-- Type: application/octet-stream, Size: 7905 bytes --]

diff --git a/recipes/mono/mono-2.4.2.3/mono-cross-compilation-for-amd64.patch b/recipes/mono/mono-2.4.2.3/mono-cross-compilation-for-amd64.patch
index e8d91da..3cf5473 100644
--- a/recipes/mono/mono-2.4.2.3/mono-cross-compilation-for-amd64.patch
+++ b/recipes/mono/mono-2.4.2.3/mono-cross-compilation-for-amd64.patch
@@ -1,23 +1,166 @@
-This is a workaround for bug 515050
 https://bugzilla.novell.com/show_bug.cgi?id=515050#c2
 
-Real fix is something else which did not make into 2.4.2
+Backport following fixes from mono-2-4 branch
 http://lists.ximian.com/pipermail/mono-patches/2009-June/151386.html
 http://lists.ximian.com/pipermail/mono-patches/2009-June/151387.html
 
-but we are ok with workaround its safe.
-
 -Khem
 Index: mono-2.4.2.3/mono/mini/genmdesc.pl
 ===================================================================
---- mono-2.4.2.3.orig/mono/mini/genmdesc.pl	2009-10-15 18:21:45.000000000 -0700
-+++ mono-2.4.2.3/mono/mini/genmdesc.pl	2009-10-15 18:22:19.000000000 -0700
-@@ -47,7 +47,7 @@ sub load_opcodes
- 		$arch_define = "TARGET_ARM";
- 	}
- 		
--	$cpp .= " -D$arch_define $srcdir/mini-ops.h|";
-+	$cpp .= " -DMONO_ARCH_SUPPORT_SIMD_INTRINSICS -D$arch_define $srcdir/mini-ops.h|";
+--- mono-2.4.2.3.orig/mono/mini/genmdesc.pl	2009-11-04 15:36:19.000000000 -0800
++++ mono-2.4.2.3/mono/mini/genmdesc.pl	2009-11-04 15:36:27.000000000 -0800
+@@ -51,7 +51,7 @@ sub load_opcodes
  	#print "Running: $cpp\n";
  	open (OPS, $cpp) || die "Cannot execute cpp: $!";
  	while (<OPS>) {
+-		next unless /MINI_OP\s*\(\s*(\S+?)\s*,\s*"(.*?)"/;
++		next unless /MINI_OP3?\s*\(\s*(\S+?)\s*,\s*"(.*?)"/;
+ 		my ($sym, $name) = ($1, $2);
+ 		push @opcodes, [$sym, $name];
+ 		$table{$name} = {num => $i, name => $name};
+@@ -159,16 +159,19 @@ sub build_table {
+ }
+ 
+ sub usage {
+-	die "genmdesc.pl arch srcdir desc output name\n";
++	die "genmdesc.pl arch srcdir output name desc [desc2 ...]\n";
+ }
+ 
+ my $arch = shift || usage ();
+ my $srcdir = shift || usage ();
+-my $file = shift || usage ();
+ my $output = shift || usage ();
+ my $name = shift || usage ();
++usage () unless @ARGV;
++my @files = @ARGV;
+ 
+ load_opcodes ($srcdir, $arch);
+-load_file ($file);
++foreach my $file (@files) {
++	load_file ($file);
++}
+ build_table ($output, $name);
+ 
+Index: mono-2.4.2.3/mono/mini/Makefile.am
+===================================================================
+--- mono-2.4.2.3.orig/mono/mini/Makefile.am	2009-11-04 15:30:20.000000000 -0800
++++ mono-2.4.2.3/mono/mini/Makefile.am	2009-11-04 15:36:27.000000000 -0800
+@@ -422,40 +422,40 @@ GENMDESC_PRG=./genmdesc
+ endif !CROSS_COMPILING
+ 
+ cpu-x86.h: cpu-x86.md genmdesc$(EXEEXT)
+-	$(GENMDESC_PRG) $(srcdir)/cpu-x86.md cpu-x86.h x86_desc
++	$(GENMDESC_PRG) cpu-x86.h x86_desc $(srcdir)/cpu-x86.md
+ 
+ cpu-amd64.h: cpu-amd64.md genmdesc$(EXEEXT)
+-	$(GENMDESC_PRG) $(srcdir)/cpu-amd64.md cpu-amd64.h amd64_desc
++	$(GENMDESC_PRG) cpu-amd64.h amd64_desc $(srcdir)/cpu-amd64.md
+ 
+ cpu-ppc.h: cpu-ppc.md genmdesc$(EXEEXT)
+-	$(GENMDESC_PRG) $(srcdir)/cpu-ppc.md cpu-ppc.h ppcg4
++	$(GENMDESC_PRG) cpu-ppc.h ppcg4 $(srcdir)/cpu-ppc.md
+ 
+ cpu-ppc64.h: cpu-ppc64.md genmdesc$(EXEEXT)
+-	$(GENMDESC_PRG) $(srcdir)/cpu-ppc64.md cpu-ppc64.h ppc64_cpu_desc
++	$(GENMDESC_PRG) cpu-ppc64.h ppc64_cpu_desc $(srcdir)/cpu-ppc64.md
+ 
+ cpu-arm.h: cpu-arm.md genmdesc$(EXEEXT)
+-	$(GENMDESC_PRG) $(srcdir)/cpu-arm.md cpu-arm.h arm_cpu_desc
++	$(GENMDESC_PRG) cpu-arm.h arm_cpu_desc $(srcdir)/cpu-arm.md
+ 
+ cpu-sparc.h: cpu-sparc.md genmdesc$(EXEEXT)
+-	$(GENMDESC_PRG) $(srcdir)/cpu-sparc.md cpu-sparc.h sparc_desc
++	$(GENMDESC_PRG) cpu-sparc.h sparc_desc $(srcdir)/cpu-sparc.md
+ 
+ cpu-s390.h: cpu-s390.md genmdesc$(EXEEXT)
+-	$(GENMDESC_PRG) $(srcdir)/cpu-s390.md cpu-s390.h s390_cpu_desc
++	$(GENMDESC_PRG) cpu-s390.h s390_cpu_desc $(srcdir)/cpu-s390.md
+ 
+ cpu-s390x.h: cpu-s390x.md genmdesc$(EXEEXT)
+-	$(GENMDESC_PRG) $(srcdir)/cpu-s390x.md cpu-s390x.h s390x_cpu_desc
++	$(GENMDESC_PRG) cpu-s390x.h s390x_cpu_desc $(srcdir)/cpu-s390x.md
+ 
+ cpu-ia64.h: cpu-ia64.md genmdesc$(EXEEXT)
+-	$(GENMDESC_PRG) $(srcdir)/cpu-ia64.md cpu-ia64.h ia64_desc
++	$(GENMDESC_PRG) cpu-ia64.h ia64_desc $(srcdir)/cpu-ia64.md
+ 
+ cpu-alpha.h: cpu-alpha.md genmdesc$(EXEEXT)
+-	$(GENMDESC_PRG) $(srcdir)/cpu-alpha.md cpu-alpha.h alpha_desc
++	$(GENMDESC_PRG) cpu-alpha.h alpha_desc $(srcdir)/cpu-alpha.md
+ 
+ cpu-hppa.h: cpu-hppa.md genmdesc$(EXEEXT)
+-	$(GENMDESC_PRG) $(srcdir)/cpu-hppa.md cpu-hppa.h hppa_desc
++	$(GENMDESC_PRG) cpu-hppa.h hppa_desc $(srcdir)/cpu-hppa.md
+ 
+ cpu-mips.h: cpu-mips.md genmdesc$(EXEEXT)
+-	$(GENMDESC_PRG) $(srcdir)/cpu-mips.md cpu-mips.h mips_desc
++	$(GENMDESC_PRG) cpu-mips.h mips_desc $(srcdir)/cpu-mips.md
+ 
+ testi: mono test.exe
+ 	$(RUNTIME) -v -v --ncompile 1 --compile Test:$(mtest) test.exe
+Index: mono-2.4.2.3/mono/mini/genmdesc.c
+===================================================================
+--- mono-2.4.2.3.orig/mono/mini/genmdesc.c	2009-11-04 15:35:34.000000000 -0800
++++ mono-2.4.2.3/mono/mini/genmdesc.c	2009-11-04 15:36:27.000000000 -0800
+@@ -211,21 +211,20 @@ int 
+ main (int argc, char* argv [])
+ {
+ 	init_table ();
+-	switch (argc) {
+-	case 2:
++	if (argc == 2) {
+ 		/* useful to get a new file when some opcodes are added: looses the comments, though */
+ 		load_file (argv [1]);
+ 		dump ();
+-		break;
+-	case 4:
+-		load_file (argv [1]);
+-		build_table (argv [2], argv [3]);
+-		break;
+-	default:
++	} else if (argc < 4) {
+ 		g_print ("Usage: genmdesc arguments\n");
+-		g_print ("\tgenmdesc desc             Output to stdout the description file.\n");
+-		g_print ("\tgenmdesc desc output name Write to output the description in a table named 'name'.\n");
++		g_print ("\tgenmdesc desc                        Output to stdout the description file.\n");
++		g_print ("\tgenmdesc output name desc [desc1...] Write to output the description in a table named 'name'.\n");
+ 		return 1;
++	} else {
++		int i;
++		for (i = 3; i < argc; ++i)
++			load_file (argv [i]);
++		build_table (argv [1], argv [2]);
+ 	}
+ 	return 0;
+ }
+Index: mono-2.4.2.3/mono/mini/mini-ops.h
+===================================================================
+--- mono-2.4.2.3.orig/mono/mini/mini-ops.h	2009-11-04 15:35:34.000000000 -0800
++++ mono-2.4.2.3/mono/mini/mini-ops.h	2009-11-04 15:38:37.000000000 -0800
+@@ -72,7 +72,7 @@ MINI_OP(OP_STOREI8_MEMBASE_REG, "storei8
+ MINI_OP(OP_STORER4_MEMBASE_REG, "storer4_membase_reg", IREG, FREG, NONE)
+ MINI_OP(OP_STORER8_MEMBASE_REG, "storer8_membase_reg", IREG, FREG, NONE)
+ 
+-#ifdef MONO_ARCH_SUPPORT_SIMD_INTRINSICS
++#if defined(TARGET_X86) || defined(TARGET_AMD64)
+ MINI_OP(OP_STOREX_MEMBASE_REG, "storex_membase_reg", IREG, XREG, NONE)
+ MINI_OP(OP_STOREX_ALIGNED_MEMBASE_REG,     "storex_aligned_membase_reg", IREG, XREG, NONE)
+ MINI_OP(OP_STOREX_NTA_MEMBASE_REG,     "storex_nta_membase_reg", IREG, XREG, NONE)
+@@ -100,7 +100,7 @@ MINI_OP(OP_LOADR8_MEMBASE,"loadr8_membas
+ 
+ MINI_OP(OP_LOADX_MEMBASE, 			"loadx_membase", XREG, IREG, NONE)
+ 
+-#ifdef MONO_ARCH_SUPPORT_SIMD_INTRINSICS
++#if defined(TARGET_X86) || defined(TARGET_AMD64)
+ MINI_OP(OP_LOADX_ALIGNED_MEMBASE,  "loadx_aligned_membase", XREG, IREG, NONE)
+ #endif
+ 
+@@ -578,7 +578,7 @@ MINI_OP(OP_NOT_NULL, "not_null", NONE, I
+ 
+ /* SIMD opcodes. */
+ 
+-#ifdef MONO_ARCH_SUPPORT_SIMD_INTRINSICS
++#if defined(TARGET_X86) || defined(TARGET_AMD64)
+ 
+ MINI_OP(OP_ADDPS, "addps", XREG, XREG, XREG)
+ MINI_OP(OP_DIVPS, "divps", XREG, XREG, XREG)
diff --git a/recipes/mono/mono_2.4.2.3.bb b/recipes/mono/mono_2.4.2.3.bb
index 5e659f3..c6e3665 100644
--- a/recipes/mono/mono_2.4.2.3.bb
+++ b/recipes/mono/mono_2.4.2.3.bb
@@ -2,7 +2,7 @@ require mono-${PV}.inc
 
 DEPENDS = "mono-native mono-mcs-intermediate glib-2.0 perl-native"
 
-PR = "${INC_PR}.1"
+PR = "${INC_PR}.2"
 
 # mono makes use of non-thumb-compatible inline asm.
 ARM_INSTRUCTION_SET = "arm"

Re: Mono

[Hodgson, Simon]

_______________________________________
From: openembedded-devel-bounces@lists.openembedded.org [openembedded-devel-bounces@lists.openembedded.org] On Behalf Of Khem Raj [raj.khem@gmail.com]
Sent: 05 November 2009 00:06
To: openembedded-devel@lists.openembedded.org
Subject: Re: [oe] Mono

On Wed, Nov 4, 2009 at 12:11 PM, Hodgson, Simon
<Simon.Hodgson@nortechonline.co.uk> wrote:
> I've built Mono for OpenEmbedded to run on an arm processor. But when I try to run just a simple Hello World test I get:
>
> mini-codegen.c:1073:mono_local_regalloc: assertion failed: (ins->dreg != -1)
>
> Does anyone have any suggestion as to what I might be doing wrong?
>
> mono -V produces:
> root@at91sam9g20ek:~# mono -V
> Mono JIT compiler version 2.4.2.3 (tarball Tue Nov  3 03:08:58 EST 2009)
> Copyright (C) 2002-2008 Novell, Inc and Contributors. www.mono-project.com
>        TLS:           normal
>        GC:            Included Boehm (with typed GC)
>        SIGSEGV:       normal
>        Notifications: epoll
>        Architecture:  armel,vfp
>        Disabled:      none
>
> Is there anymore info perhaps I can provide?

Can you apply the attached patch then rebuild mono and retry your test.

Thanks

-Khem


Sorry for being a bit dense, but where should I apply this? It appears to patch the receipe and the mono source, is that correct? Do I need to run patch twice in each directory?

Thanks


Simon

Re: Mono

[Khem Raj]

On Thu, Nov 5, 2009 at 1:20 PM, Hodgson, Simon
<Simon.Hodgson@nortechonline.co.uk> wrote:
> _______________________________________
> From: openembedded-devel-bounces@lists.openembedded.org [openembedded-devel-bounces@lists.openembedded.org] On Behalf Of Khem Raj [raj.khem@gmail.com]
> Sent: 05 November 2009 00:06
> To: openembedded-devel@lists.openembedded.org
> Subject: Re: [oe] Mono
>
> On Wed, Nov 4, 2009 at 12:11 PM, Hodgson, Simon
> <Simon.Hodgson@nortechonline.co.uk> wrote:
>> I've built Mono for OpenEmbedded to run on an arm processor. But when I try to run just a simple Hello World test I get:
>>
>> mini-codegen.c:1073:mono_local_regalloc: assertion failed: (ins->dreg != -1)
>>
>> Does anyone have any suggestion as to what I might be doing wrong?
>>
>> mono -V produces:
>> root@at91sam9g20ek:~# mono -V
>> Mono JIT compiler version 2.4.2.3 (tarball Tue Nov  3 03:08:58 EST 2009)
>> Copyright (C) 2002-2008 Novell, Inc and Contributors. www.mono-project.com
>>        TLS:           normal
>>        GC:            Included Boehm (with typed GC)
>>        SIGSEGV:       normal
>>        Notifications: epoll
>>        Architecture:  armel,vfp
>>        Disabled:      none
>>
>> Is there anymore info perhaps I can provide?
>
> Can you apply the attached patch then rebuild mono and retry your test.
>
> Thanks
>
> -Khem
>
>
> Sorry for being a bit dense, but where should I apply this? It appears to patch the receipe and the mono source, is that correct? Do I need to run patch twice in each directory?

you can cd into (top of your oe source tree) say openembedded/ then do

patch -p1 < this.patch

>
> Thanks
>
>
> Simon
> _______________________________________________
> Openembedded-devel mailing list
> Openembedded-devel@lists.openembedded.org
> http://lists.linuxtogo.org/cgi-bin/mailman/listinfo/openembedded-devel
>