Bug #4411: Context saving issue on PPC64

• Status: Closed
• Assignee: -
• Target version: -
• ruby -v: -

[a3li (Alex Legler)]

=begin
Hi,

please see this bug report with a stacktrace and gdb info: https://bugzilla.redhat.com/show_bug.cgi?id=628715
It also includes a proposed patch which I'd like to have considered for inclusion. The issue is still there in p334.

Thanks!
=end

[naruse (Yui NARUSE)]

• Priority changed from Normal to 3

=begin
We can't see brewweb.devel.redhat.com.
=end

[a3li (Alex Legler)]

=begin
I cannot access it either, it seems to be redhat-internal.

In case you were looking for the patch, it is here: https://bugzilla.redhat.com/attachment.cgi?id=445801
=end

[nobu (Nobuyoshi Nakada)]

• ruby -v changed from ruby 1.8.7 (2010-06-23 patchlevel 299) [powerpc64-linux] to -

=begin
Hi,

At Tue, 22 Feb 2011 15:33:59 +0900,
Alex Legler wrote in [ruby-core:35333]:

In case you were looking for the patch, it is here: https://bugzilla.redhat.com/attachment.cgi?id=445801

It's gccish code.

• (((j)->status)?:(just_before_setjmp)), \
  

--
Nobu Nakada
=end

[nobu (Nobuyoshi Nakada)]

=begin
Hi,

At Tue, 22 Feb 2011 15:33:59 +0900,
Alex Legler wrote in [ruby-core:35333]:

In case you were looking for the patch, it is here: https://bugzilla.redhat.com/attachment.cgi?id=445801

It's gccish code.

• (((j)->status)?:(just_before_setjmp)), \
  

--
Nobu Nakada
=end

[nobu (Nobuyoshi Nakada)]

=begin
Hi,

At Tue, 22 Feb 2011 15:33:59 +0900,
Alex Legler wrote in [ruby-core:35333]:

In case you were looking for the patch, it is here: https://bugzilla.redhat.com/attachment.cgi?id=445801

It's gccish code.

• (((j)->status)?:(just_before_setjmp)), \
  

--
Nobu Nakada
=end

It's gccish code.

• (((j)->status)?:(just_before_setjmp)), \
  

GCCish or not it's been few months with no reaction here. Vanilla ruby is therefore unusable on PPC64. Can you provide better patch or at least a suggestion how we should fix it "in a better way"?
Cheers,
Kacper Kowalik

[nobu (Nobuyoshi Nakada)]

=begin
Hi,

: At Tue, 17 May 2011 03:50:32 +0900,
: Kacper Kowalik wrote in [ruby-core:36232]:
GCCish or not it's been few months with no reaction
here. Vanilla ruby is therefore unusable on PPC64. Can you
provide better patch or at least a suggestion how we should
fix it "in a better way"?

Does this work?

diff --git a/eval.c b/eval.c
index a54fdce..e25ee24 100644
--- a/eval.c
+++ b/eval.c
@@ -188,6 +188,9 @@ int function_call_may_return_twice_false_2 = 0;
(function_call_may_return_twice_false_2 ?
setjmp(function_call_may_return_twice_jmp_buf) :
0)
+# elif defined(__ia64)
+# define JUST_BEFORE_SETJMP(extra_save, j) ((void)0)
+# define JUST_AFTER_SETJMP(extra_save, j) ((j)->status ? (void)0 : (extra_save))

elif defined(FreeBSD) && FreeBSD < 7

/*

• workaround for FreeBSD/i386 getcontext/setcontext bug.
  @@ -205,16 +208,23 @@ static int volatile freebsd_clear_carry_flag = 0;

ifndef POST_GETCONTEXT

define POST_GETCONTEXT 0

endif

+# ifndef JUST_BEFORE_SETJMP
+# define JUST_BEFORE_SETJMP(extra_save, j) (extra_save)
+# endif
+# ifndef JUST_AFTER_SETJMP
+# define JUST_AFTER_SETJMP(extra_save, j) ((void)0)
+# endif

define ruby_longjmp(env, val) rb_jump_context(env, val)

-# define ruby_setjmp(just_before_setjmp, j) ((j)->status = 0, \

• (just_before_setjmp), \
  

+# define ruby_setjmp(extra_save, j) ((j)->status = 0, \

• JUST_BEFORE_SETJMP(extra_save, j), \
  PRE_GETCONTEXT, \
  getcontext(&(j)->context), \
  POST_GETCONTEXT, \
  

• JUST_AFTER_SETJMP(extra_save, j), \
  (j)->status)
  

#else
-# define ruby_setjmp(just_before_setjmp, env) \

• ((just_before_setjmp), RUBY_SETJMP(env))
  

+# define ruby_setjmp(extra_save, env) \

• ((extra_save), RUBY_SETJMP(env))
  

define ruby_longjmp(env,val) RUBY_LONGJMP(env,val)

ifdef CYGWIN

ifndef _setjmp

--- Nobu Nakada
=end

Nobuyoshi Nakada wrote:

Does this work?
Yes, thank you! But of course you need s/__ia64/PPC64/ following line:

+# elif defined(__ia64)
+# define JUST_BEFORE_SETJMP(extra_save, j) ((void)0)
Cheers,
Kacper Kowalik

Could this patch be applied to 1.8 branch?

[nobu (Nobuyoshi Nakada)]

Hi,

At Wed, 18 May 2011 02:23:35 +0900,
Kacper Kowalik wrote in [ruby-core:36277]:

Yes, thank you! But of course you need s/__ia64/PPC64/ following line:

Thank you, I also noticed it just after the post.
Is it all upper case?

--
Nobu Nakada

Nobuyoshi Nakada wrote:

Hi,

At Wed, 18 May 2011 02:23:35 +0900,
Kacper Kowalik wrote in [ruby-core:36277]:

Yes, thank you! But of course you need s/__ia64/PPC64/ following line:

Thank you, I also noticed it just after the post.
Is it all upper case?
Yes,

$ echo | gcc -dM -E - | grep -i ppc
#define _ARCH_PPCGR 1
#define PPC64 1
#define _ARCH_PPC 1
#define PPC 1
#define _ARCH_PPC64 1

[nobu (Nobuyoshi Nakada)]

• Description updated (diff)
• Status changed from Open to Closed
• % Done changed from 0 to 100

Fixed at r32542