Mirror of the gdb-patches mailing list
 help / color / mirror / Atom feed
From: Tom Tromey <tromey@redhat.com>
To: Yao Qi <yao@codesourcery.com>
Cc: <gdb-patches@sourceware.org>
Subject: Re: RFC: close-on-exec internal file descriptors
Date: Wed, 24 Apr 2013 02:45:00 -0000	[thread overview]
Message-ID: <871ua1qmc3.fsf@fleche.redhat.com> (raw)
In-Reply-To: <5175DF3E.10800@codesourcery.com> (Yao Qi's message of "Tue, 23	Apr 2013 09:09:18 +0800")

>>>>> "Yao" == Yao Qi <yao@codesourcery.com> writes:

Yao> This patch causes a build regression with mingw32 cross gcc, reported
Yao> here http://sourceware.org/ml/gdb-patches/2013-04/msg00682.html and I
Yao> can reproduce it on my fedora 16:

Thanks for the note.  I'm Sorry about the breakage.

Yao> Your patch doesn't mention much on why not to check
Yao> sys/socket.h. Probably we still need to check it.

I looked and other places use USE_WIN32API.

I am checking in the appended.  It fixes the mingw build for me using
the mingw cross compilers that come with Fedora 18.  I also built and
regression tested this natively.

I did consider gnulib but it seems we already have a reasonably working
solution; and using it just for mode_t (see the patch) seemed like
overkill.

Tom

2013-04-23  Tom Tromey  <tromey@redhat.com>

	* common/filestuff.c: Check USE_WIN32API before including
	sys/socket.h.
	(HAVE_F_GETFD): New define.
	(mark_cloexec): Check HAVE_F_GETFD.
	(gdb_open_cloexec): Change 'mode' to unsigned long.
	(gdb_socketpair_cloexec): Check HAVE_SOCKETPAIR.
	(gdb_pipe_cloexec): Check HAVE_PIPE.
	* common/filestuff.h (gdb_open_cloexec): Change 'mode' to unsigned
	long.

Index: common/filestuff.c
===================================================================
RCS file: /cvs/src/src/gdb/common/filestuff.c,v
retrieving revision 1.1
diff -u -r1.1 filestuff.c
--- common/filestuff.c	22 Apr 2013 16:46:15 -0000	1.1
+++ common/filestuff.c	23 Apr 2013 15:43:46 -0000
@@ -28,10 +28,18 @@
 #include <string.h>
 #include <fcntl.h>
 #include <unistd.h>
-#include <sys/socket.h>
 #include <sys/types.h>
 #include <sys/stat.h>
 
+#ifdef USE_WIN32API
+#include <winsock2.h>
+#include <windows.h>
+#else
+#include <sys/socket.h>
+/* Define HAVE_F_GETFD if we plan to use F_GETFD.  */
+#define HAVE_F_GETFD F_GETFD
+#endif
+
 #ifdef HAVE_SYS_RESOURCE_H
 #include <sys/resource.h>
 #endif /* HAVE_SYS_RESOURCE_H */
@@ -215,6 +223,7 @@
 static void
 mark_cloexec (int fd)
 {
+#ifdef HAVE_F_GETFD
   int old = fcntl (fd, F_GETFD, 0);
 
   if (old != -1)
@@ -229,6 +238,7 @@
 	    trust_o_cloexec = -1;
 	}
     }
+#endif /* HAVE_F_GETFD */
 }
 
 /* Depending on TRUST_O_CLOEXEC, mark FD as close-on-exec.  */
@@ -254,7 +264,7 @@
 /* See filestuff.h.  */
 
 int
-gdb_open_cloexec (const char *filename, int flags, mode_t mode)
+gdb_open_cloexec (const char *filename, int flags, unsigned long mode)
 {
   int fd = open (filename, flags | O_CLOEXEC, mode);
 
@@ -303,6 +313,7 @@
 int
 gdb_socketpair_cloexec (int namespace, int style, int protocol, int filedes[2])
 {
+#ifdef HAVE_SOCKETPAIR
   int result = socketpair (namespace, style | SOCK_CLOEXEC, protocol, filedes);
 
   if (result != -1)
@@ -312,6 +323,9 @@
     }
 
   return result;
+#else
+  gdb_assert_not_reached (_("socketpair not available on this host"));
+#endif
 }
 
 /* See filestuff.h.  */
@@ -342,13 +356,17 @@
       maybe_mark_cloexec (filedes[1]);
     }
 #else
+#ifdef HAVE_PIPE
   result = pipe (filedes);
   if (result != -1)
     {
       mark_cloexec (filedes[0]);
       mark_cloexec (filedes[1]);
     }
-#endif
+#else /* HAVE_PIPE */
+  gdb_assert_not_reached (_("pipe not available on this host"));
+#endif /* HAVE_PIPE */
+#endif /* HAVE_PIPE2 */
 
   return result;
 }
Index: common/filestuff.h
===================================================================
RCS file: /cvs/src/src/gdb/common/filestuff.h,v
retrieving revision 1.1
diff -u -r1.1 filestuff.h
--- common/filestuff.h	22 Apr 2013 16:46:15 -0000	1.1
+++ common/filestuff.h	23 Apr 2013 15:43:46 -0000
@@ -33,7 +33,8 @@
 /* Like 'open', but ensures that the returned file descriptor has the
    close-on-exec flag set.  */
 
-extern int gdb_open_cloexec (const char *filename, int flags, mode_t mode);
+extern int gdb_open_cloexec (const char *filename, int flags,
+			     /* mode_t */ unsigned long mode);
 
 /* Like 'fopen', but ensures that the returned file descriptor has the
    close-on-exec flag set.  */


  reply	other threads:[~2013-04-23 15:49 UTC|newest]

Thread overview: 13+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2012-12-18 17:45 Tom Tromey
2012-12-18 22:12 ` Andreas Schwab
2012-12-19 14:53   ` Tom Tromey
2013-01-03 16:17     ` Tom Tromey
2013-04-22 23:55       ` Tom Tromey
2013-04-23  1:51         ` Tom Tromey
2013-05-01 14:47         ` Joel Brobecker
2013-05-06 18:53           ` Tom Tromey
2013-05-07  6:46             ` Joel Brobecker
2013-05-10 17:00               ` Tom Tromey
2013-04-23 15:45 ` Yao Qi
2013-04-24  2:45   ` Tom Tromey [this message]
2013-04-24 20:35     ` Jan Kratochvil

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=871ua1qmc3.fsf@fleche.redhat.com \
    --to=tromey@redhat.com \
    --cc=gdb-patches@sourceware.org \
    --cc=yao@codesourcery.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox