[Cluster-devel] [PATCH] cleanup build system a bit.

[Cluster-devel] [PATCH] cleanup build system a bit.

[Fabio Massimo Di Nitto]

Hi all,

the patch in attachment if the first big pass to cleanup the entire build
system on CVS HEAD as of 24-Apr-2007. Sorry if it's a bz2 file but otherwise it
won't pass the mailing list filtering limit of 100KB and it's one of those
patches that either you take in full or not (except a bunch of small details).

 193 files changed, 1927 insertions(+), 8223 deletions(-)

This patch does:

 - centralize all the configure scripts into the top level one:
   - all configurable options are now available as such with details and help.
   - no more need to run configure N times for each subdir.
   - better autodetection for kernel_src.
   - add tons of other options such as alternate CC or LD_FLAGS override and
     more.
   - allow override of release_major and release_minor.
 - centralize all the different subproject/make/* files in the top level:
   - expanded and integraded all the defines.mk.input into a clean one.
 - centralize copyright.cf file at the toplevel. config/copyright.cf
   and make it a default include directory for all projects so that there
   is no more need to specify it manually.
 - centralize scripts at the top level as appropriate.
 - remove all the projects/configure projects/make/* projects/scripts/*
 - change top level Makefile includes:
   - implement several targets and add appropriate dependencies.
   - split between kernel and userland targets.
 - changes to all subprojects Makefiles:
   - make use of the top level defines.mk.
   - make use of centralized scripts.
   - make use of all configurable options.
   - cleanup TARGET= OBJS= CFLAGS= LDFLAGS= usage and make it consistent
     across the entire tree. (see below for a more detailed explanation)
   - fix make targets for libs.
   - call linker with proper options and fix static/dynamic linking.
   - attempt to use the use the same targets all over for consistency.
   - fix install/uninstall targets.
   - convert some toplevel project Makefile to be just "passtrough".
     This is no different from what it was before except it takes 4 lines
     instead of N duplicates lines. It will change in the next patch set
     to integrate proper dependency calculation. See for example
     cman/Makefile.
 - probably more...

The way in which I rewrote the Makefile follows some kind of template.
For (or what i believe to be) better readability, I used this schema:

TARGET= foo
OBJS= foo.o

CFLAGS += -Ddefinition
CFLAGS += compiler options such as -O2 or -Wall
CFLAGS += system includes or configured includes
CFLAGS += includes coming from inside the project for ex: -I../include

similar order applies to LDFLAGS.

While at it i also fixed these 4 details:

cluster/fence/agents/xvm/vm_states.c: fix include path for openais.

cluster/gfs/gfs_quota/layout.c: fix include path to be consistent with the
                                rest of the tree.
(see: https://www.redhat.com/archives/cluster-devel/2007-April/msg00115.html)

cluster/group/lib/libgroup.c: add missing include. Spotted after fixing the
                              Makefile and triggered by adding -Wall.

cluster/rgmanager/src/clulib/clulog.c: remove unrequired include.

TODO:

somebody from RH should please verify that config/copyright.cf is correct.
The oldest RH copyright is from 2004 and i updated from 2006 to 2007.

top level configure:
        - fix hardencoded use of xml2-config and make it optional

top level makefile:
        - verify latest_tags: srpms: tarballs: targets.
        I have personally never used them and for what i can tell they haven't
        been in used for ages.

verify rgmanager/src/Makefile for check: target

verify rgmanager/src/daemon/Makefile dtest and rg_test OBJ list.
        - see OBJS2 and OBJS3 build of resrule*.o.

While I did pay an awful amount of attention not to break anything, there is
a chance that something might have gone wrong.

Please double check and apply.

Thanks
Fabio

-- 
I'm going to make him an offer he can't refuse.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: cleanup_configure.bz2
Type: application/x-bzip
Size: 32594 bytes
Desc: not available
URL: <http://listman.redhat.com/archives/cluster-devel/attachments/20070424/b42b5aa8/attachment.bin>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 890 bytes
Desc: OpenPGP digital signature
URL: <http://listman.redhat.com/archives/cluster-devel/attachments/20070424/b42b5aa8/attachment.sig>

[Cluster-devel] [PATCH] cleanup build system a bit.

[Fabio Massimo Di Nitto]

Just spotten on IRC by Patrick.

Please apply after the previous patch.

Thanks
Fabio

-- 
I'm going to make him an offer he can't refuse.
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: group.diff
URL: <http://listman.redhat.com/archives/cluster-devel/attachments/20070424/e3384a22/attachment.ksh>

[Cluster-devel] [PATCH] cleanup build system a bit.

[Lon Hohberger]

On Tue, Apr 24, 2007 at 03:34:04PM +0200, Fabio Massimo Di Nitto wrote:
> Just spotten on IRC by Patrick.
> 
> Please apply after the previous patch.
> 
> Thanks
> Fabio
> 
> -- 
> I'm going to make him an offer he can't refuse.

> diff -Naurd cluster.orig2/group/daemon/Makefile cluster/group/daemon/Makefile
> --- cluster.orig2/group/daemon/Makefile	2007-04-24 11:15:43.000000000 +0200
> +++ cluster/group/daemon/Makefile	2007-04-24 15:32:07.000000000 +0200
> @@ -24,7 +24,7 @@
>  CFLAGS += -I${cmanincdir}
>  CFLAGS += -I. -I../include/ -I../lib/
>  
> -LDFLAGS += -L${cmanlibdir} -L${openaislibdir} -L{libdir} -lcman -lcpg
> +LDFLAGS += -L${cmanlibdir} -L${openaislibdir} -L${libdir} -lcman -lcpg
>  
>  all: ${TARGET}
>  

FWIW, this is an overhaul not a cleanup. ;)

I'm fine with it, but I want more people to comment.

(expecting minor bugs to be found, but easy to fix)

-- Lon

-- 
Lon Hohberger - Software Engineer - Red Hat, Inc.

[Cluster-devel] [PATCH] cleanup build system a bit.

[Fabio Massimo Di Nitto]

Lon Hohberger wrote:

> 
> FWIW, this is an overhaul not a cleanup. ;)

details :P

> 
> I'm fine with it, but I want more people to comment.
> 
> (expecting minor bugs to be found, but easy to fix)

sure.. i am of course dealing to fix them as soon as they arise.

Thanks
Fabio

-- 
I'm going to make him an offer he can't refuse.

[Cluster-devel] [PATCH] cleanup build system a bit.

[Chris Feist]

Fabio Massimo Di Nitto wrote:
> TODO:
> 
> somebody from RH should please verify that config/copyright.cf is correct.
> The oldest RH copyright is from 2004 and i updated from 2006 to 2007.

Looks good

> top level makefile:
>         - verify latest_tags: srpms: tarballs: targets.
>         I have personally never used them and for what i can tell they haven't
>         been in used for ages.

latest_tags, build_srpms, tarballs are from the old, old build system, they can 
definitely be removed.

I think this patch is something that we can commit to the -HEAD branch and start 
using for FC8 and beyond.  We shouldn't have a problem patching up any minor 
issues that crop up.  I vote for committing this patch.

Thanks,
Chris

[Cluster-devel] [PATCH] cleanup build system a bit.

[Fabio Massimo Di Nitto]

Hi Chris,

Chris Feist wrote:
> Fabio Massimo Di Nitto wrote:

>> top level makefile:
>>         - verify latest_tags: srpms: tarballs: targets.
>>         I have personally never used them and for what i can tell they haven't
>>         been in used for ages.
> 
> latest_tags, build_srpms, tarballs are from the old, old build system, they can 
> definitely be removed.
> 
> I think this patch is something that we can commit to the -HEAD branch and start 
> using for FC8 and beyond.  We shouldn't have a problem patching up any minor 
> issues that crop up.  I vote for committing this patch.

Thanks for the feedback. I will provide a patch to cleanup those once/if the
main patch will make it in CVS.

Cheers
Fabio

-- 
I'm going to make him an offer he can't refuse.

[Cluster-devel] [PATCH] cleanup build system a bit.

[Fabio Massimo Di Nitto]

Fabio Massimo Di Nitto wrote:
> Hi all,

There is a new patch in attachment that address (including a rediff towards new
Makefile in CVS HEAD):

> verify rgmanager/src/daemon/Makefile dtest and rg_test OBJ list.
>         - see OBJS2 and OBJS3 build of resrule*.o.

what Chris Feist reported:

> latest_tags, build_srpms, tarballs are from the old, old build system, they can 
> definitely be removed.

and the typo reported by Patrick on IRC.

No other changes have been done.

Thanks
Fabio

-- 
I'm going to make him an offer he can't refuse.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: bigfat.diff.bz2
Type: application/x-bzip
Size: 32743 bytes
Desc: not available
URL: <http://listman.redhat.com/archives/cluster-devel/attachments/20070426/d27e2d7c/attachment.bin>

[Cluster-devel] [PATCH] cleanup build system a bit.

[Fabio Massimo Di Nitto]

Fabio Massimo Di Nitto wrote:

> No other changes have been done.

So far I have seen a few "thumbs up" on the patch. If nobody objects I would
like to commit it in CVS HEAD together with a bunch of other small fixes I have
in the queue (patches will follow and are based on top of this one).

Thanks
Fabio

-- 
I'm going to make him an offer he can't refuse.

[Cluster-devel] [PATCH 001/008] Fix configure help typo

[Fabio Massimo Di Nitto]

-- 
I'm going to make him an offer he can't refuse.
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: 002_fix_configure_typo.dpatch
URL: <http://listman.redhat.com/archives/cluster-devel/attachments/20070427/8143a019/attachment.ksh>

[Cluster-devel] [PATCH 002/008]

[Fabio Massimo Di Nitto]

Fix configure parameter check for libexecdir.

-- 
I'm going to make him an offer he can't refuse.
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: 003_fix_missing_configure_check.dpatch
URL: <http://listman.redhat.com/archives/cluster-devel/attachments/20070427/dc9a7c43/attachment.ksh>

[Cluster-devel] [PATCH 003/008] Fix --release_* check

[Fabio Massimo Di Nitto]

-- 
I'm going to make him an offer he can't refuse.
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: 004_fix_release_checking.dpatch
URL: <http://listman.redhat.com/archives/cluster-devel/attachments/20070427/c8381fd7/attachment.ksh>

[Cluster-devel] [PATCH 004/008] gnbd_export: convert harcoded DEFAULT_GETUID to build flag

[Fabio Massimo Di Nitto]

-- 
I'm going to make him an offer he can't refuse.
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: 005_gnbd_make_default_getuid_configurable.dpatch
URL: <http://listman.redhat.com/archives/cluster-devel/attachments/20070427/6137109d/attachment.ksh>

[Cluster-devel] [PATCH 005/008] Fix userland: target depends

[Fabio Massimo Di Nitto]

-- 
I'm going to make him an offer he can't refuse.
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: 006_fix_top_makefile_dependencies.dpatch
URL: <http://listman.redhat.com/archives/cluster-devel/attachments/20070427/822641aa/attachment.ksh>

[Cluster-devel] [PATCH 006/008] Fix make all target in passtrough Makefiles and make them more consistent

[Fabio Massimo Di Nitto]

-- 
I'm going to make him an offer he can't refuse.
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: 007_fix_passtrough_makefiles.dpatch
URL: <http://listman.redhat.com/archives/cluster-devel/attachments/20070427/f99e3de7/attachment.ksh>

[Cluster-devel] [PATCH 007/008] Readd -Werror to rgmanager dropped by mistake in the big cleanup.

[Fabio Massimo Di Nitto]

-- 
I'm going to make him an offer he can't refuse.
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: 008_rgmanager_readd_werror.dpatch
URL: <http://listman.redhat.com/archives/cluster-devel/attachments/20070427/2ec0c28e/attachment.ksh>

[Cluster-devel] [PATCH 008/008] Readd SHAREDIR define to rgmanager/src/daemons/Makefile

[Fabio Massimo Di Nitto]

-- 
I'm going to make him an offer he can't refuse.
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: 009_rgmanager_needs_sharedir.dpatch
URL: <http://listman.redhat.com/archives/cluster-devel/attachments/20070427/bcfb4247/attachment.ksh>

[Cluster-devel] [PATCH] cleanup build system a bit.

[Robert Peterson]

Fabio Massimo Di Nitto wrote:
>> I think this patch is something that we can commit to the -HEAD branch and start 
>> using for FC8 and beyond.  We shouldn't have a problem patching up any minor 
>> issues that crop up.  I vote for committing this patch.

Hi Fabio,

This looks like a good patch to me.  I've been too busy to look at it in
any great detail, but I did look at the highlights.  I have a minor concern
about the Makefile in gfs2/mkfs: IIRC, it looked like the default "Make" 
(all) specified only mkfs.gfs2 as the target, whereas we need to be sure to
that gfs2_jadd and gfs2_grow get built as well.  Er, well, I guess you don't
have gfs2_grow yet (but I do).  Maybe it's already assured with some magic 
and trickery, as it often happens in many Makefiles, but that will be apparent
soon enough if/when I get a chance to try it.

As with any change with so many things affected, there might be some minor 
glitches or gotchas that we'll find, but I think we can work around those 
as we encounter them.  Good job!

Regards,

Bob Peterson
Red Hat Cluster Suite

[Cluster-devel] [PATCH] cleanup build system a bit.

[Fabio Massimo Di Nitto]

Robert Peterson wrote:
> Fabio Massimo Di Nitto wrote:
>>> I think this patch is something that we can commit to the -HEAD branch and start 
>>> using for FC8 and beyond.  We shouldn't have a problem patching up any minor 
>>> issues that crop up.  I vote for committing this patch.
> 
> Hi Fabio,
> 
> This looks like a good patch to me.  I've been too busy to look at it in
> any great detail, but I did look at the highlights. 

great thanks.

> I have a minor concern
> about the Makefile in gfs2/mkfs: IIRC, it looked like the default "Make" 
> (all) specified only mkfs.gfs2 as the target, whereas we need to be sure to
> that gfs2_jadd and gfs2_grow get built as well.

Right.. patch in attachment to turn the Makefile in line with the others.

>  Er, well, I guess you don't
> have gfs2_grow yet (but I do).

well i can't really cope with stuff that's not in CVS yet :))) really.. even
when i try hard ;)

>  Maybe it's already assured with some magic 
> and trickery, as it often happens in many Makefiles, but that will be apparent
> soon enough if/when I get a chance to try it.

Once you commit gfs2_grow in CVS HEAD, i can add it to the Makefile.

> As with any change with so many things affected, there might be some minor 
> glitches or gotchas that we'll find, but I think we can work around those 
> as we encounter them.

I got a CVS access, so i can fix most of them once they are reported tho I was
asked to always ask for patch review before applying, so it's just a matter for
somebody to ACK.

>  Good job!

Thanks
Fabio

PS I plan to commit sometime on monday EU time. If you can add gfs2_grow i will
merge the Makefiles at the same time.

-- 
I'm going to make him an offer he can't refuse.
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: 010_gfs2_mkfs_cleanup.dpatch
URL: <http://listman.redhat.com/archives/cluster-devel/attachments/20070427/186d3702/attachment.ksh>