NetBSD Problem Report #49989

From www@NetBSD.org  Sat Jun 20 15:01:18 2015
Return-Path: <www@NetBSD.org>
Received: from mail.netbsd.org (mail.netbsd.org [149.20.53.66])
	(using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256 bits))
	(Client CN "mail.netbsd.org", Issuer "Postmaster NetBSD.org" (verified OK))
	by mollari.NetBSD.org (Postfix) with ESMTPS id 588CDA654F
	for <gnats-bugs@gnats.NetBSD.org>; Sat, 20 Jun 2015 15:01:18 +0000 (UTC)
Message-Id: <20150620150117.16A6CA65E4@mollari.NetBSD.org>
Date: Sat, 20 Jun 2015 15:01:17 +0000 (UTC)
From: pooka@iki.fi
Reply-To: pooka@iki.fi
To: gnats-bugs@NetBSD.org
Subject: libc++ fails to build with g++ 5.1 due to libpthread volatiles
X-Send-Pr-Version: www-1.0

>Number:         49989
>Category:       lib
>Synopsis:       libc++ fails to build with g++ 5.1 due to libpthread volatiles
>Confidential:   no
>Severity:       serious
>Priority:       medium
>Responsible:    lib-bug-people
>State:          closed
>Class:          sw-bug
>Submitter-Id:   net
>Arrival-Date:   Sat Jun 20 15:05:00 +0000 2015
>Closed-Date:    Thu Aug 27 12:35:04 +0000 2015
>Last-Modified:  Thu Aug 27 12:35:04 +0000 2015
>Originator:     Antti Kantee
>Release:        some future NetBSD
>Organization:
>Environment:
>Description:
libc++ fails to build with g++ 5.1 due to the following types of errors:

=== snip ===
libc++/dist/libcxx/include/__mutex_base:32:69: error: temporary of non-literal type '__pthread_mutex_st' in a constant expression
      constexpr mutex() _NOEXCEPT : __m_(PTHREAD_MUTEX_INITIALIZER) {}
[...]
include/pthread_types.h:101:8: note: '__pthread_mutex_st' is not literal because:
 struct __pthread_mutex_st {
        ^
include/pthread_types.h:103:17: note:   non-static data member '__pthread_mutex_st::ptm_errorcheck' has volatile type
  pthread_spin_t ptm_errorcheck;
=== snip ===

For a full build log including the full g++ error output, see:
https://gist.githubusercontent.com/anonymous/566a26a930fa2d902894/raw/6824c3c138580bff415caddec3bdf1062d560a31/build.log

(via rumprun issue #33)
>How-To-Repeat:
On a(n x86) system with g++ 5.1, git clone repo.rumpkernel.org/rumprun and run "./build-rr.sh hw".
>Fix:
No idea, I don't speak C++.  Not sure if this is a libpthread problem or a libc++ problem.  (it might even be a g++ problem, in which case it's not a NetBSD bug and this report can be closed, but my money is on the first two and hence submitting it here)

>Release-Note:

>Audit-Trail:
From: christos@zoulas.com (Christos Zoulas)
To: gnats-bugs@NetBSD.org, lib-bug-people@netbsd.org, 
	gnats-admin@netbsd.org, netbsd-bugs@netbsd.org
Cc: 
Subject: Re: lib/49989: libc++ fails to build with g++ 5.1 due to libpthread volatiles
Date: Sat, 20 Jun 2015 12:04:52 -0400

 On Jun 20,  3:05pm, pooka@iki.fi (pooka@iki.fi) wrote:
 -- Subject: lib/49989: libc++ fails to build with g++ 5.1 due to libpthread v

 Index: pthread_types.h
 ===================================================================
 RCS file: /cvsroot/src/lib/libpthread/pthread_types.h,v
 retrieving revision 1.13
 diff -u -u -r1.13 pthread_types.h
 --- pthread_types.h	2 Aug 2008 19:46:30 -0000	1.13
 +++ pthread_types.h	20 Jun 2015 16:03:04 -0000
 @@ -84,6 +84,23 @@
  };

  /*
 + * libc++ expects to be using PTHREAD_MUTEX_INITIALIZER as a member
 + * initializer. This does not work for volatile types. Since c++
 + * does not touch the guts of those types, we redefine them as non-volatile
 + */
 +#ifdef __cplusplus
 +# ifdef __CPU_SIMPLE_LOCK_PAD
 +#  define __pthread_mutex_spin_t unsigned char
 +# else
 +#  define __pthread_mutex_spin_t unsigned int
 +# endif
 +# define __pthread_mutex_volatile
 +#else
 +# define __pthread_mutex_spin_t pthread_spin_t
 +# define __pthread_mutex_volatile volatile
 +#endif
 +
 +/*
   * ptm_owner is the actual lock field which is locked via CAS operation.
   * This structure's layout is designed to compatible with the previous
   * version used in SA pthreads.
 @@ -100,16 +117,16 @@
  #endif
  struct	__pthread_mutex_st {
  	unsigned int	ptm_magic;
 -	pthread_spin_t	ptm_errorcheck;
 +	__pthread_mutex_spin_t	ptm_errorcheck;
  #ifdef __CPU_SIMPLE_LOCK_PAD
  	uint8_t		ptm_pad1[3];
  #endif
 -	pthread_spin_t	ptm_interlock;	/* unused - backwards compat */
 +	__pthread_mutex_spin_t	ptm_interlock;	/* unused - backwards compat */
  #ifdef __CPU_SIMPLE_LOCK_PAD
  	uint8_t		ptm_pad2[3];
  #endif
 -	volatile pthread_t ptm_owner;
 -	pthread_t * volatile ptm_waiters;
 +	__pthread_mutex_volatile pthread_t ptm_owner;
 +	pthread_t * __pthread_mutex_volatile ptm_waiters;
  	unsigned int	ptm_recursed;
  	void		*ptm_spare2;	/* unused - backwards compat */
  };

From: Antti Kantee <pooka@iki.fi>
To: gnats-bugs@NetBSD.org, lib-bug-people@netbsd.org, 
 gnats-admin@netbsd.org, netbsd-bugs@netbsd.org
Cc: 
Subject: Re: lib/49989: libc++ fails to build with g++ 5.1 due to libpthread
 volatiles
Date: Sat, 20 Jun 2015 16:12:15 +0000

 Thanks, Christos!

 I'll ask the submitter to test.

 On 20/06/15 16:05, Christos Zoulas wrote:
 >   Index: pthread_types.h
 >   ===================================================================
 >   RCS file: /cvsroot/src/lib/libpthread/pthread_types.h,v
 >   retrieving revision 1.13
 >   diff -u -u -r1.13 pthread_types.h
 >   --- pthread_types.h	2 Aug 2008 19:46:30 -0000	1.13
 >   +++ pthread_types.h	20 Jun 2015 16:03:04 -0000
 >   @@ -84,6 +84,23 @@
 >    };
 >
 >    /*
 >   + * libc++ expects to be using PTHREAD_MUTEX_INITIALIZER as a member
 >   + * initializer. This does not work for volatile types. Since c++
 >   + * does not touch the guts of those types, we redefine them as non-volatile
 >   + */
 >   +#ifdef __cplusplus
 >   +# ifdef __CPU_SIMPLE_LOCK_PAD
 >   +#  define __pthread_mutex_spin_t unsigned char
 >   +# else
 >   +#  define __pthread_mutex_spin_t unsigned int
 >   +# endif
 >   +# define __pthread_mutex_volatile
 >   +#else
 >   +# define __pthread_mutex_spin_t pthread_spin_t
 >   +# define __pthread_mutex_volatile volatile
 >   +#endif
 >   +
 >   +/*
 >     * ptm_owner is the actual lock field which is locked via CAS operation.
 >     * This structure's layout is designed to compatible with the previous
 >     * version used in SA pthreads.
 >   @@ -100,16 +117,16 @@
 >    #endif
 >    struct	__pthread_mutex_st {
 >    	unsigned int	ptm_magic;
 >   -	pthread_spin_t	ptm_errorcheck;
 >   +	__pthread_mutex_spin_t	ptm_errorcheck;
 >    #ifdef __CPU_SIMPLE_LOCK_PAD
 >    	uint8_t		ptm_pad1[3];
 >    #endif
 >   -	pthread_spin_t	ptm_interlock;	/* unused - backwards compat */
 >   +	__pthread_mutex_spin_t	ptm_interlock;	/* unused - backwards compat */
 >    #ifdef __CPU_SIMPLE_LOCK_PAD
 >    	uint8_t		ptm_pad2[3];
 >    #endif
 >   -	volatile pthread_t ptm_owner;
 >   -	pthread_t * volatile ptm_waiters;
 >   +	__pthread_mutex_volatile pthread_t ptm_owner;
 >   +	pthread_t * __pthread_mutex_volatile ptm_waiters;
 >    	unsigned int	ptm_recursed;
 >    	void		*ptm_spare2;	/* unused - backwards compat */
 >    };
 >
 >

From: "Brian G." <drakain@gmail.com>
To: gnats-bugs@netbsd.org
Cc: 
Subject: Re: lib/49989
Date: Sat, 20 Jun 2015 18:47:53 -0400

 --001a113ce222c326e60518fad434
 Content-Type: text/plain; charset=UTF-8

 This patch fixes the problem for me when I also replace pthread_spin_t with
 __pthread_mutex_spin_t inside __pthread_cond_st and rwlock_t.

 --001a113ce222c326e60518fad434
 Content-Type: text/html; charset=UTF-8

 <p dir="ltr">This patch fixes the problem for me when I also replace pthread_spin_t with __pthread_mutex_spin_t inside __pthread_cond_st and rwlock_t.<br></p>

 --001a113ce222c326e60518fad434--

From: Joerg Sonnenberger <joerg@britannica.bec.de>
To: gnats-bugs@NetBSD.org
Cc: 
Subject: Re: lib/49989: libc++ fails to build with g++ 5.1 due to libpthread
 volatiles
Date: Tue, 23 Jun 2015 12:33:51 +0200

 On Sat, Jun 20, 2015 at 03:05:00PM +0000, pooka@iki.fi wrote:
 > === snip ===
 > libc++/dist/libcxx/include/__mutex_base:32:69: error: temporary of non-literal type '__pthread_mutex_st' in a constant expression
 >       constexpr mutex() _NOEXCEPT : __m_(PTHREAD_MUTEX_INITIALIZER) {}
 > [...]
 > include/pthread_types.h:101:8: note: '__pthread_mutex_st' is not literal because:
 >  struct __pthread_mutex_st {
 >         ^
 > include/pthread_types.h:103:17: note:   non-static data member '__pthread_mutex_st::ptm_errorcheck' has volatile type
 >   pthread_spin_t ptm_errorcheck;

 This is likely an interaction of GCC not believing this to be a system
 hader and a bug in the C++11 specs when it comes to volatile in nested
 elements.

 Joerg

From: Antti Kantee <pooka@iki.fi>
To: gnats-bugs@NetBSD.org, lib-bug-people@netbsd.org, 
 gnats-admin@netbsd.org, netbsd-bugs@netbsd.org
Cc: 
Subject: Re: lib/49989: libc++ fails to build with g++ 5.1 due to libpthread
 volatiles
Date: Tue, 23 Jun 2015 11:56:15 +0000

 On 23/06/15 10:35, Joerg Sonnenberger wrote:
 > The following reply was made to PR lib/49989; it has been noted by GNATS.
 >
 > From: Joerg Sonnenberger <joerg@britannica.bec.de>
 > To: gnats-bugs@NetBSD.org
 > Cc:
 > Subject: Re: lib/49989: libc++ fails to build with g++ 5.1 due to libpthread
 >   volatiles
 > Date: Tue, 23 Jun 2015 12:33:51 +0200
 >
 >   On Sat, Jun 20, 2015 at 03:05:00PM +0000, pooka@iki.fi wrote:
 >   > === snip ===
 >   > libc++/dist/libcxx/include/__mutex_base:32:69: error: temporary of non-literal type '__pthread_mutex_st' in a constant expression
 >   >       constexpr mutex() _NOEXCEPT : __m_(PTHREAD_MUTEX_INITIALIZER) {}
 >   > [...]
 >   > include/pthread_types.h:101:8: note: '__pthread_mutex_st' is not literal because:
 >   >  struct __pthread_mutex_st {
 >   >         ^
 >   > include/pthread_types.h:103:17: note:   non-static data member '__pthread_mutex_st::ptm_errorcheck' has volatile type
 >   >   pthread_spin_t ptm_errorcheck;
 >
 >   This is likely an interaction of GCC not believing this to be a system
 >   hader and a bug in the C++11 specs when it comes to volatile in nested
 >   elements.

 Ok, so we'll use Christos' patch with the small additions as per Brian's 
 test.

From: Martin Husemann <martin@duskware.de>
To: Antti Kantee <pooka@iki.fi>
Cc: gnats-bugs@NetBSD.org
Subject: Re: lib/49989: libc++ fails to build with g++ 5.1 due to libpthread volatiles
Date: Tue, 23 Jun 2015 15:19:40 +0200

 On Tue, Jun 23, 2015 at 11:56:15AM +0000, Antti Kantee wrote:
 > Ok, so we'll use Christos' patch with the small additions as per Brian's 
 > test.

 Isn't this actually a gcc bug?

 Martin

State-Changed-From-To: open->closed
State-Changed-By: pooka@NetBSD.org
State-Changed-When: Fri, 26 Jun 2015 01:33:35 +0000
State-Changed-Why:
fixed


From: "Antti Kantee" <pooka@netbsd.org>
To: gnats-bugs@gnats.NetBSD.org
Cc: 
Subject: PR/49989 CVS commit: src/lib/libpthread
Date: Fri, 26 Jun 2015 01:33:09 +0000

 Module Name:	src
 Committed By:	pooka
 Date:		Fri Jun 26 01:33:09 UTC 2015

 Modified Files:
 	src/lib/libpthread: pthread_types.h

 Log Message:
 C++ (namely libc++) expects to be using PTHREAD_FOO_INITIALIZER as a
 member initializer. This does not work for volatile types. Since C++
 does not touch the guts of those types, redefine them as non-volatile.

 Fixes libc++ compilation with g++ 5.1, as reported in PR lib/49989.


 To generate a diff of this commit:
 cvs rdiff -u -r1.13 -r1.14 src/lib/libpthread/pthread_types.h

 Please note that diffs are not public domain; they are subject to the
 copyright notices on the relevant files.

From: "Antti Kantee" <pooka@netbsd.org>
To: gnats-bugs@gnats.NetBSD.org
Cc: 
Subject: PR/49989 CVS commit: src/doc
Date: Fri, 26 Jun 2015 10:25:41 +0000

 Module Name:	src
 Committed By:	pooka
 Date:		Fri Jun 26 10:25:41 UTC 2015

 Modified Files:
 	src/doc: HACKS

 Log Message:
 document g++ 5.1 + libpthread volatile workaround (PR lib/49989)


 To generate a diff of this commit:
 cvs rdiff -u -r1.159 -r1.160 src/doc/HACKS

 Please note that diffs are not public domain; they are subject to the
 copyright notices on the relevant files.

State-Changed-From-To: closed->open
State-Changed-By: pooka@NetBSD.org
State-Changed-When: Fri, 26 Jun 2015 13:47:03 +0000
State-Changed-Why:
well that didn't work.  ports are too divergent for us to be able to
make a simple deduction on what __cpu_simplelock_t really looks like.


From: matthew green <mrg@eterna.com.au>
To: gnats-bugs@NetBSD.org
Cc: lib-bug-people@netbsd.org, netbsd-bugs@netbsd.org,
    gnats-admin@netbsd.org, pooka@NetBSD.org, pooka@iki.fi
Subject: re: lib/49989 (libc++ fails to build with g++ 5.1 due to libpthread volatiles)
Date: Sun, 28 Jun 2015 22:11:26 +1000

 > well that didn't work.  ports are too divergent for us to be able to
 > make a simple deduction on what __cpu_simplelock_t really looks like.

 ... the whole point of opaque types is to be opaque?  :-)

State-Changed-From-To: open->closed
State-Changed-By: pooka@NetBSD.org
State-Changed-When: Thu, 27 Aug 2015 12:35:04 +0000
State-Changed-Why:
fixed (again, hope it stays fixed this time)


>Unformatted:

Home	PR Database Search