On Thu, Feb 19, 2009 at 02:07:27AM +0000, Kyle McMartin wrote:
Modified Files: kernel.spec Log Message: go away docs. we hates you.
And here's a patch to turn them back on once per -rc. I think. My shell isn't what it used to be.
The idea is we don't build docs other than on the first rebase to an -rc. Which is what the else bit accomplished. I think.
Index: Makefile =================================================================== RCS file: /cvs/pkgs/rpms/kernel/devel/Makefile,v retrieving revision 1.97 diff -u -p -r1.97 Makefile --- Makefile 5 Feb 2009 18:55:52 -0000 1.97 +++ Makefile 19 Feb 2009 02:28:32 -0000 @@ -114,6 +114,7 @@ release: @perl -pi -e 's/CONFIG_NR_CPUS=512/CONFIG_NR_CPUS=64/' config-x86_64-generic
@perl -pi -e 's/^%define debugbuildsenabled 0/%define debugbuildsenabled 1/' kernel.spec + @perl -pi -e 's/^%define with_doc 0/#% define with_doc 0/' kernel.spec
rhel: @perl -pi -e 's/# CONFIG_PPC_64K_PAGES is not set/CONFIG_PPC_64K_PAGES=y/' config-powerpc64 Index: scripts/rebase.sh =================================================================== RCS file: /cvs/pkgs/rpms/kernel/devel/scripts/rebase.sh,v retrieving revision 1.24 diff -u -p -r1.24 rebase.sh --- scripts/rebase.sh 14 Jan 2009 00:55:01 -0000 1.24 +++ scripts/rebase.sh 19 Feb 2009 02:28:32 -0000 @@ -72,14 +72,24 @@ echo "NEW kernel is $NEW BASE=$NEWBASE if [ "$OLDRC" -eq 0 -a "$OLDGIT" -eq 0 -a "$OLDGIT" -ne "$NEWGIT" ]; then echo "Rebasing from a stable release to a new git snapshot" perl -p -i -e 's/^%define\ released_kernel\ 1/%define\ released_kernel\ 0/' kernel.spec + perl -p -i -e 's/^%define\ with_doc\ 0/#%\ define\ with_doc\ 0/' kernel.spec # force these to zero in this case, they may not have been when we rebased to stable perl -p -i -e 's/^%define\ rcrev.*/%define\ rcrev\ 0/' kernel.spec perl -p -i -e 's/^%define\ gitrev.*/%define\ gitrev\ 0/' kernel.spec fi
+# make sure we build docs at least once per -rc kernel, shut it off otherwise +if [ "$OLDRC" -ne 0 -a "$NEWRC" -gt "$OLDRC" ]; then + perl -p -i -e 's/^%define\ with_doc\ 0/#%\ define\ with_doc\ 0/' kernel.spec +else if [ "$NEWRC" -eq "$OLDRC" -a "$NEWGIT" -gt "$OLDGIT" ]; then + # common case, same -rc, new -git, make sure docs are off. + perl -p -i -e 's/^#%\ define\ with_doc\ 0/%define\ with_doc\ 0/' kernel.spec +endif + if [ "$NEWRC" -eq 0 -a "$NEWGIT" -eq 0 ]; then echo "Rebasing from -rc to final release." perl -p -i -e 's/^%define\ released_kernel\ 0/%define\ released_kernel\ 1/' kernel.spec + perl -p -i -e 's/^%define\ with_doc\ 0/#%\ define\ with_doc\ 0/' kernel.spec export OLD_TARBALL_BASE=$(($OLDBASE-1)) perl -p -i -e 's/^%define\ base_sublevel\ $ENV{OLD_TARBALL_BASE}/%define\ base_sublevel\ $ENV{NEWBASE}/' kernel.spec perl -p -i -e 's/^%define\ rcrev.*/%define\ rcrev\ 0/' kernel.spec
Kyle McMartin (kyle@mcmartin.ca) said:
On Thu, Feb 19, 2009 at 02:07:27AM +0000, Kyle McMartin wrote:
Modified Files: kernel.spec Log Message: go away docs. we hates you.
And here's a patch to turn them back on once per -rc. I think. My shell isn't what it used to be.
The idea is we don't build docs other than on the first rebase to an -rc. Which is what the else bit accomplished. I think.
The compose process (for rawhide/updates) only pulls the latest build for each source package; ergo, this won't work as far as havong -docs always available to install.
Bill
On Thu, Feb 19, 2009 at 09:53:57AM -0500, Bill Nottingham wrote:
The compose process (for rawhide/updates) only pulls the latest build for each source package; ergo, this won't work as far as havong -docs always available to install.
Well, that's broken. But honestly I suspect nobody actually cares. And if they do, I'm sure the F-10 package is still installable.
Kyle McMartin (kyle@infradead.org) said:
On Thu, Feb 19, 2009 at 09:53:57AM -0500, Bill Nottingham wrote:
The compose process (for rawhide/updates) only pulls the latest build for each source package; ergo, this won't work as far as havong -docs always available to install.
Well, that's broken. But honestly I suspect nobody actually cares. And if they do, I'm sure the F-10 package is still installable.
Why not just make it conditional on whatver the release/non-debug/etc. switch is?
Bill
On Thu, Feb 19, 2009 at 10:30:56AM -0500, Bill Nottingham wrote:
Kyle McMartin (kyle@infradead.org) said:
On Thu, Feb 19, 2009 at 09:53:57AM -0500, Bill Nottingham wrote:
The compose process (for rawhide/updates) only pulls the latest build for each source package; ergo, this won't work as far as havong -docs always available to install.
Well, that's broken. But honestly I suspect nobody actually cares. And if they do, I'm sure the F-10 package is still installable.
Why not just make it conditional on whatver the release/non-debug/etc. switch is?
I don't care if it gets /composed/ the goal of building it every once and a while is to make sure it actually does build. We're pretty much the only people shipping top of tree code in a distro...
Granted, special case this release since we're shipping the same thing in F-10 anyway.
On Thu, 19 Feb 2009 10:33:03 -0500 Kyle McMartin kyle@infradead.org wrote:
On Thu, Feb 19, 2009 at 10:30:56AM -0500, Bill Nottingham wrote:
Kyle McMartin (kyle@infradead.org) said:
On Thu, Feb 19, 2009 at 09:53:57AM -0500, Bill Nottingham wrote:
The compose process (for rawhide/updates) only pulls the latest build for each source package; ergo, this won't work as far as havong -docs always available to install.
Well, that's broken. But honestly I suspect nobody actually cares. And if they do, I'm sure the F-10 package is still installable.
Why not just make it conditional on whatver the release/non-debug/etc. switch is?
I don't care if it gets /composed/ the goal of building it every once and a while is to make sure it actually does build. We're pretty much the only people shipping top of tree code in a distro...
But once the release switch gets flipped we should always build it.
kernel@lists.fedoraproject.org