bug#39363: emacs-git version: pthread_setname_np on NetBSD

bug#39363: emacs-git version: pthread_setname_np on NetBSD

[Thomas Klausner]

[-- Attachment #1: Type: text/plain, Size: 493 bytes --]

Hi!

Recently, emacs from git stopped compiling on NetBSD because it
started using pthread_setname_np. AFAIK, there is no commonly agreed
upon standard for this function, and NetBSD's uses three arguments, see
https://netbsd.gw.com/cgi-bin/man-cgi?pthread_setname_np++NetBSD-current

The attached patch makes emacs compile again (on NetBSD-9.99.43/amd64)
but configure should probably be taught to look for that version of
pthread_setname_np instead of the #ifdef __NetBSD__.

Thanks,
 Thomas

[-- Attachment #2: patch-src_systhread.c --]
[-- Type: text/plain, Size: 450 bytes --]

$NetBSD$

Adapt pthread_setname_np calling convention for NetBSD.

--- src/systhread.c.orig	2020-01-30 22:42:52.049505198 +0000
+++ src/systhread.c
@@ -217,7 +217,11 @@ sys_thread_set_name (const char *name)
  #ifdef HAVE_PTHREAD_SETNAME_NP_1ARG
   pthread_setname_np (p_name);
  #else
+  #ifdef __NetBSD__
+  pthread_setname_np (pthread_self (), "%s", p_name);
+  #else
   pthread_setname_np (pthread_self (), p_name);
+  #endif
  #endif
 #endif
 }

bug#39363: emacs-git version: pthread_setname_np on NetBSD

[Eli Zaretskii]

> Date: Fri, 31 Jan 2020 00:02:13 +0100
> From: Thomas Klausner <wiz@NetBSD.org>
> 
> Recently, emacs from git stopped compiling on NetBSD because it
> started using pthread_setname_np. AFAIK, there is no commonly agreed
> upon standard for this function, and NetBSD's uses three arguments, see
> https://netbsd.gw.com/cgi-bin/man-cgi?pthread_setname_np++NetBSD-current
> 
> The attached patch makes emacs compile again (on NetBSD-9.99.43/amd64)
> but configure should probably be taught to look for that version of
> pthread_setname_np instead of the #ifdef __NetBSD__.

I think we want instead to modify the configure-time test to include
the possibility of 3-argument pthread_setname_np.

Thanks.

bug#39363: emacs-git version: pthread_setname_np on NetBSD

[Robert Pluim]

>>>>> On Fri, 31 Jan 2020 09:48:13 +0200, Eli Zaretskii <eliz@gnu.org> said:

    >> Date: Fri, 31 Jan 2020 00:02:13 +0100
    >> From: Thomas Klausner <wiz@NetBSD.org>
    >> 
    >> Recently, emacs from git stopped compiling on NetBSD because it
    >> started using pthread_setname_np. AFAIK, there is no commonly agreed
    >> upon standard for this function, and NetBSD's uses three arguments, see
    >> https://netbsd.gw.com/cgi-bin/man-cgi?pthread_setname_np++NetBSD-current
    >> 
    >> The attached patch makes emacs compile again (on NetBSD-9.99.43/amd64)
    >> but configure should probably be taught to look for that version of
    >> pthread_setname_np instead of the #ifdef __NetBSD__.

    Eli> I think we want instead to modify the configure-time test to include
    Eli> the possibility of 3-argument pthread_setname_np.

Untested patch (I donʼt have any NetBSD machines).

diff --git a/configure.ac b/configure.ac
index f604acb694..1d922d9c02 100644
--- a/configure.ac
+++ b/configure.ac
@@ -4192,6 +4192,21 @@ AC_DEFUN
     AC_DEFINE(
       HAVE_PTHREAD_SETNAME_NP_1ARG, 1,
       [Define to 1 if pthread_setname_np takes a single argument.])
+  else
+    AC_CACHE_CHECK(
+     [whether pthread_setname_np takes three arguments],
+     [emacs_cv_pthread_setname_np_3arg],
+     [AC_COMPILE_IFELSE(
+       [AC_LANG_PROGRAM(
+         [[#include <pthread.h>]],
+         [[pthread_setname_np (0, "%s", "a");]])],
+       [emacs_cv_pthread_setname_np_3arg=yes],
+       [emacs_cv_pthread_setname_np_3arg=no])])
+     if test "$emacs_cv_pthread_setname_np_3arg" = "yes"; then
+       AC_DEFINE(
+         HAVE_PTHREAD_SETNAME_NP_3ARG, 1,
+         [Define to 1 if pthread_setname_np takes three arguments.])
+     fi
   fi
 fi
 
diff --git a/src/systhread.c b/src/systhread.c
index c649ae853a..0d600d6895 100644
--- a/src/systhread.c
+++ b/src/systhread.c
@@ -214,11 +214,13 @@ #define TASK_COMM_LEN 16
   char p_name[TASK_COMM_LEN];
   strncpy (p_name, name, TASK_COMM_LEN - 1);
   p_name[TASK_COMM_LEN - 1] = '\0';
- #ifdef HAVE_PTHREAD_SETNAME_NP_1ARG
+# ifdef HAVE_PTHREAD_SETNAME_NP_1ARG
   pthread_setname_np (p_name);
- #else
+# elif defined HAVE_PTHREAD_SETNAME_NP_3ARG
+  pthread_setname_np (pthread_self (), "%s", p_name);
+# else
   pthread_setname_np (pthread_self (), p_name);
- #endif
+# endif
 #endif
 }
 

bug#39363: emacs-git version: pthread_setname_np on NetBSD

[Thomas Klausner]

On Fri, Jan 31, 2020 at 10:02:50AM +0100, Robert Pluim wrote:
>     Eli> I think we want instead to modify the configure-time test to include
>     Eli> the possibility of 3-argument pthread_setname_np.
> 
> Untested patch (I donʼt have any NetBSD machines).

Works for me. Thank you!
 Thomas

bug#39363: emacs-git version: pthread_setname_np on NetBSD

[Robert Pluim]

>>>>> On Mon, 03 Feb 2020 10:38:38 +0100, Robert Pluim <rpluim@gmail.com> said:

>>>>> On Fri, 31 Jan 2020 18:09:57 +0100, Thomas Klausner <wiz@NetBSD.org> said:
    Thomas> On Fri, Jan 31, 2020 at 10:02:50AM +0100, Robert Pluim wrote:
    Eli> I think we want instead to modify the configure-time test to include
    Eli> the possibility of 3-argument pthread_setname_np.
    >>> 
    >>> Untested patch (I donʼt have any NetBSD machines).

    Thomas> Works for me. Thank you!

Thanks for testing. Eli, emacs-27 I presume?

Robert

bug#39363: emacs-git version: pthread_setname_np on NetBSD

[Eli Zaretskii]

> From: Robert Pluim <rpluim@gmail.com>
> Date: Mon, 03 Feb 2020 12:37:02 +0100
> Cc: 39363@debbugs.gnu.org
> 
>     >>> Untested patch (I donʼt have any NetBSD machines).
> 
>     Thomas> Works for me. Thank you!
> 
> Thanks for testing. Eli, emacs-27 I presume?

Yes, thanks.

bug#39363: emacs-git version: pthread_setname_np on NetBSD

[Robert Pluim]

>>>>> On Mon, 03 Feb 2020 17:35:18 +0200, Eli Zaretskii <eliz@gnu.org> said:

    >> From: Robert Pluim <rpluim@gmail.com>
    >> Date: Mon, 03 Feb 2020 12:37:02 +0100
    >> Cc: 39363@debbugs.gnu.org
    >> 
    >> >>> Untested patch (I donʼt have any NetBSD machines).
    >> 
    Thomas> Works for me. Thank you!
    >> 
    >> Thanks for testing. Eli, emacs-27 I presume?

    Eli> Yes, thanks.

Done. Closing bug.

Robert