Lustre devel - Dec 2006 - [Bug 10947] Require SLES 10 support for Lustre

Please don''t reply to lustre-devel. Instead, comment in Bugzilla by
using the following link:
https://bugzilla.lustre.org/show_bug.cgi?id=10947

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Attachment #8989|review?(green@clusterfs.com)|review+
               Flag|                            |


(From update of attachment 8989)
Please land to b1_5

Please don''t reply to lustre-devel. Instead, comment in Bugzilla by
using the following link:
https://bugzilla.lustre.org/show_bug.cgi?id=10947



(In reply to comment #48)
> (From update of attachment 8989)
> Please land to b1_5
done.
>

Please don''t reply to lustre-devel. Instead, comment in Bugzilla by
using the following link:
https://bugzilla.lustre.org/show_bug.cgi?id=10947



Created an attachment (id=9252)
Please don''t reply to lustre-devel. Instead, comment in Bugzilla by
using the following link:
 --> (https://bugzilla.lustre.org/attachment.cgi?id=9252&action=view)
Configure checks to determine return type of readlink()

This has been committed to b1_4_sles10 (and, unfortunately, HEAD for our
libsysio).  The configure checks are necessary to build in a real SLES10
environment.

Please don''t reply to lustre-devel. Instead, comment in Bugzilla by
using the following link:
https://bugzilla.lustre.org/show_bug.cgi?id=10947



(From update of attachment 9252)
I sent this patch to Lee@Sandia and he included it in his tree.
>Index: libsysio/include/sysio.h
>==================================================================>RCS
file: /cvsroot/cfs/libsysio/include/sysio.h,v
>retrieving revision 1.3
>retrieving revision 1.8
>diff -u -r1.3 -r1.8
>--- libsysio/include/sysio.h	14 Dec 2006 00:49:06 -0000	1.3
>+++ libsysio/include/sysio.h	28 Dec 2006 04:53:40 -0000	1.8
>@@ -247,7 +247,11 @@
> #endif
> extern int SYSIO_INTERFACE_NAME(rmdir)(const char *path);
> extern int SYSIO_INTERFACE_NAME(symlink)(const char *path1, const char
*path2);
>+#ifdef HAVE_POSIX_1003_READLINK
>+extern ssize_t SYSIO_INTERFACE_NAME(readlink)(const char *path,
>+#else
> extern int SYSIO_INTERFACE_NAME(readlink)(const char *path,
>+#endif
> 				char *buf,
> 				size_t bufsiz);
> extern int SYSIO_INTERFACE_NAME(link)(const char *oldpath, const char
*newpath);

I still oppose this part below.
Because while it works for us, what it actually means is every application that
is to use libsysio, must have this configure check. (and probably every
application using liblustre too?)
What should be done instead, is some way to include libsysio''s config.h
in
sysio.h above.
>Index: lustre/autoconf/lustre-core.m4
>==================================================================>RCS
file: /cvsroot/cfs/lustre-core/autoconf/lustre-core.m4,v
>retrieving revision 1.1.4.37.2.4
>retrieving revision 1.1.4.37.2.9
>diff -u -r1.1.4.37.2.4 -r1.1.4.37.2.9
>--- lustre/autoconf/lustre-core.m4	16 Nov 2006 19:21:39 -0000	1.1.4.37.2.4
>+++ lustre/autoconf/lustre-core.m4	28 Dec 2006 07:03:14 -0000	1.1.4.37.2.9
>@@ -446,6 +446,23 @@
> ])
> ])
> 
>+#
>+# LC_READLINK_SSIZE_T
>+#
>+AC_DEFUN([LC_READLINK_SSIZE_T],
>+[AC_MSG_CHECKING([if readlink returns ssize_t])
>+AC_TRY_COMPILE([
>+	#include <unistd.h>
>+],[
>+	ssize_t readlink(const char *, char *, size_t);
>+],[
>+	AC_MSG_RESULT([yes])
>+	AC_DEFINE(HAVE_POSIX_1003_READLINK, 1, [readlink returns ssize_t])
>+],[
>+	AC_MSG_RESULT([no])
>+])
>+])
>+
> AC_DEFUN([LC_FUNC_PAGE_MAPPED],
> [AC_MSG_CHECKING([if kernel offers page_mapped])
> LB_LINUX_TRY_COMPILE([
>@@ -788,7 +805,8 @@
> AC_CHECK_HEADERS([netinet/in.h arpa/inet.h catamount/data.h])
> AC_CHECK_FUNCS([inet_ntoa])
> 
>-
>+# libsysio/src/readlink.c
>+LC_READLINK_SSIZE_T
> 
> # utils/llverfs.c
> AC_CHECK_HEADERS([ext2fs/ext2fs.h])

Please don''t reply to lustre-devel. Instead, comment in Bugzilla by
using the following link:
https://bugzilla.lustre.org/show_bug.cgi?id=10947



now depends on new bug #11562  "racer-correctness test fails on b1_4_sles10
kernel 2.6.16.21-08-sles10 on x86_64"