[Ocfs2-devel] Small fixes for ocfs2-tools 1.1.1

[Ocfs2-devel] Small fixes for ocfs2-tools 1.1.1

[Lars Marowsky-Bree]

Skipped content of type multipart/mixed-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: not available
Url : http://oss.oracle.com/pipermail/ocfs2-devel/attachments/20050820/55791fc9/attachment.bin

[Ocfs2-devel] Small fixes for ocfs2-tools 1.1.1

[Manish Singh]

On Sat, Aug 20, 2005 at 04:19:34PM +0200, Lars Marowsky-Bree wrote:
> Hi,
> 
> I'm just building ocfs2-tools 1.1.1 (bleeding edge ;-) for openSUSE.
> 
> Attached you'll find two patches.
> 
> One fixes obvious mistakes in configure.in as well as an uninitialized
> variable being dereferenced.
> 
> The second one I'm not sure about whether you want to take it upstream.
> If linking against com_err, you need to also link against pthread now.

What changed to require this? That sounds kind of broken, and isn't
reflected in upstream e2fsprogs.
 
> In our specfile, I also had to move debugfs.ocfs2 and ocfs2cdsl from
> /sbin to /usr/sbin, because they link against glib, which is not on /,
> and thus our build system doesn't like binaries on / linked against
> that.

How about moving glib to / ? ;)
 
> The third attachment are the remaining signedness warnings the compiler
> generates. I've glanced at a few, and they seem harmless (except maybe
> the one in dump.c?), but you may want to clean them up...

Yeah, gcc4 spews like mad now...

> --- configure.in	(revision 1063)
> +++ configure.in	(working copy)
> @@ -84,7 +84,7 @@
>  
>  UUID_LIBS=
>  AC_CHECK_LIB(uuid, uuid_unparse, UUID_LIBS=-luuid)
> -if test "x$COM_ERR_LIBS" = "x"; then
> +if test "x$UUID_LIBS" = "x"; then
>    AC_MSG_ERROR([Unable to find uuid library])
>  fi
>  AC_CHECK_HEADER(uuid/uuid.h, :,

Cut'n'paste screw up. Thanks for noticing, this is applied to SVN.

-Manish

[Ocfs2-devel] Small fixes for ocfs2-tools 1.1.1

[Lars Marowsky-Bree]

On 2005-08-20T13:48:13, Manish Singh <manish.singh@oracle.com> wrote:

> > The second one I'm not sure about whether you want to take it upstream.
> > If linking against com_err, you need to also link against pthread now.
> What changed to require this? That sounds kind of broken, and isn't
> reflected in upstream e2fsprogs.

I investigated in some more detail, and this seems to be our fault ;-)
We have a patch which adds mutex locking to e2fsprogs. Looks like
the libraries were made thread-safe(r). I've pinged the maintainer
if/why the patch is not upstream.

> > In our specfile, I also had to move debugfs.ocfs2 and ocfs2cdsl from
> > /sbin to /usr/sbin, because they link against glib, which is not on
> > /, and thus our build system doesn't like binaries on / linked
> > against that.
> How about moving glib to / ? ;)

Unlikely, I'm afraid ;-)


Sincerely,
    Lars Marowsky-Br?e <lmb@suse.de>

-- 
High Availability & Clustering
SUSE Labs, Research and Development
SUSE LINUX Products GmbH - A Novell Business	 -- Charles Darwin
"Ignorance more frequently begets confidence than does knowledge"

[Ocfs2-devel] Small fixes for ocfs2-tools 1.1.1

[Manish Singh]

On Sat, Aug 20, 2005 at 11:32:22PM +0200, Lars Marowsky-Bree wrote:
> On 2005-08-20T13:48:13, Manish Singh <manish.singh@oracle.com> wrote:
> 
> > > The second one I'm not sure about whether you want to take it upstream.
> > > If linking against com_err, you need to also link against pthread now.
> > What changed to require this? That sounds kind of broken, and isn't
> > reflected in upstream e2fsprogs.
> 
> I investigated in some more detail, and this seems to be our fault ;-)
> We have a patch which adds mutex locking to e2fsprogs. Looks like
> the libraries were made thread-safe(r). I've pinged the maintainer
> if/why the patch is not upstream.

Hm, then the patch needs work. com_err should be built and linked with
-pthread, so it gets a shared library dependency on libpthread, so this
becomes a non-issue. com_err.pc should also have -pthread in Cflags and
Libs.

To get it upstream, there's much more work involved, since e2fsprogs
supports more than just Linux. The incantations required to compile
against pthreads differ in subtle ways across the various platforms. 

I'd suggest reverting the change until something is worked out upstream,
so there's no gratuitous incompatibilities.

-Manish

[Ocfs2-devel] Small fixes for ocfs2-tools 1.1.1

[Joel Becker]

On Sat, Aug 20, 2005 at 04:19:34PM +0200, Lars Marowsky-Bree wrote:
> The second one I'm not sure about whether you want to take it upstream.
> If linking against com_err, you need to also link against pthread now.

	I'm so tired of arbitrary "punish the 99% of users who are sane
by pulling in threads, which are only used by 1% of folks" patches.
Then you have harder compatibility issues, and you have larger run sets,
and you have whack stuff going on when debugging...
 
> In our specfile, I also had to move debugfs.ocfs2 and ocfs2cdsl from
> /sbin to /usr/sbin, because they link against glib, which is not on /,
> and thus our build system doesn't like binaries on / linked against
> that.

	Oh, you mean distro A puts thing in different and more
surprising place than distro B?  That _never_ happens in Linux! :-)

Joel

-- 

"The only way to get rid of a temptation is to yield to it."
         - Oscar Wilde 

Joel Becker
Senior Member of Technical Staff
Oracle
E-mail: joel.becker@oracle.com
Phone: (650) 506-8127