threads / patch / 644

patchRe: [PATCH] uml: remove elf.h

Subject: Re: [PATCH] uml: remove elf.h

## tl;dr

4 messages between May 17, 2005 and May 17, 2005. Diffs are folded; open one to read it.

replies: 3people: 3as markdown or json

Andrew Morton· May 17, 2005, 21:21 UTC · lore

Linux Kernel Mailing List <linux-kernel@vger.kernel.org> wrote:

Show 20 quoted lines

>
> tree a3d85d9f43f64bbd8437c973caf98f79d95b5f3e
> parent a123edab03ac39e08c2f9cb4fc1af07e099c68bc
> author Paolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it> Tue, 17 May 2005 11:53:14 -0700
> committer Linus Torvalds <torvalds@ppc970.osdl.org> Tue, 17 May 2005 21:59:11 -0700
> 
> [PATCH] uml: remove elf.h
> 
> Actually remove elf.h in the tree.  The previous patch, due to a quilt
> bug/misuse, left it in the tree as a 0-length file, preventing the build to
> see it as missing and to generate a symlink in its place.
> 
> Signed-off-by: Paolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it>
> Signed-off-by: Andrew Morton <akpm@osdl.org>
> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
> 
>  asm-um/elf.h |    0 
>  1 files changed
> 
> Index: include/asm-um/elf.h

Hot damn, this zero-length file is hard to get rid of. I pulled Linus's tree this morning with this bizarre concoction:

	cd $GIT_TREE
	cg-pull origin
	tagsha1=$(cat .git/refs/tags/v$(kversion))
	t=$(cat-file tag $tagsha1 | head -n 1 | sed -e 's/object //')
	cg-diff -r $t -r $(cat .git/refs/heads/origin) > $PULL/linus.patch

and the resulting diff has:

Show changes to diff +0 −1

Index: include/asm-ia64/ioctl32.h
===================================================================
--- eed337ef5e9ae7d62caa84b7974a11fddc7f06e0/include/asm-ia64/ioctl32.h  (mode:100644 sha1:d0d227f45e05d23705ac849f4bd5c06a28288b58)
+++ 6bb5a1cf91bbda8308ec7e6d900cb89071907dcd/include/asm-ia64/ioctl32.h  (mode:100644 sha1:e69de29bb2d1d6434b8b29ae775ad8c2e48c5391)
@@ -1 +0,0 @@
-#include <linux/ioctl32.h>
Index: include/asm-um/elf.h
===================================================================
Index: include/asm-x86_64/apicdef.h
===================================================================

which of course doesn't remove that file at all.

And I bet that when Linus releases patch-2.6.12-rc5.gz and patch-2.6.12.gz,
they will have the same construct.  AFAICT, the patch-based people will
need to download a full new tarball to get rid of this dang file.

It all wouldn't really matter much, except apparently the mere presence of
this file breaks the UML build.

Frazzle.  Paolo, I'm almost wondering if we should change that test to also
check for a zero-length file.

Petr Baudis· May 17, 2005, 21:34 UTC · re: Andrew Morton · lore

Dear diary, on Tue, May 17, 2005 at 11:21:13PM CEST, I got a letter where Andrew Morton <akpm@osdl.org> told me that...

Show 49 quoted lines

> Linux Kernel Mailing List <linux-kernel@vger.kernel.org> wrote:
> >
> > tree a3d85d9f43f64bbd8437c973caf98f79d95b5f3e
> > parent a123edab03ac39e08c2f9cb4fc1af07e099c68bc
> > author Paolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it> Tue, 17 May 2005 11:53:14 -0700
> > committer Linus Torvalds <torvalds@ppc970.osdl.org> Tue, 17 May 2005 21:59:11 -0700
> > 
> > [PATCH] uml: remove elf.h
> > 
> > Actually remove elf.h in the tree.  The previous patch, due to a quilt
> > bug/misuse, left it in the tree as a 0-length file, preventing the build to
> > see it as missing and to generate a symlink in its place.
> > 
> > Signed-off-by: Paolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it>
> > Signed-off-by: Andrew Morton <akpm@osdl.org>
> > Signed-off-by: Linus Torvalds <torvalds@osdl.org>
> > 
> >  asm-um/elf.h |    0 
> >  1 files changed
> > 
> > Index: include/asm-um/elf.h
> 
> Hot damn, this zero-length file is hard to get rid of.  I pulled Linus's
> tree this morning with this bizarre concoction:
> 
> 	cd $GIT_TREE
> 	cg-pull origin
> 	tagsha1=$(cat .git/refs/tags/v$(kversion))
> 	t=$(cat-file tag $tagsha1 | head -n 1 | sed -e 's/object //')
> 	cg-diff -r $t -r $(cat .git/refs/heads/origin) > $PULL/linus.patch
> 
> and the resulting diff has:
> 
> Index: include/asm-ia64/ioctl32.h
> ===================================================================
> --- eed337ef5e9ae7d62caa84b7974a11fddc7f06e0/include/asm-ia64/ioctl32.h  (mode:100644 sha1:d0d227f45e05d23705ac849f4bd5c06a28288b58)
> +++ 6bb5a1cf91bbda8308ec7e6d900cb89071907dcd/include/asm-ia64/ioctl32.h  (mode:100644 sha1:e69de29bb2d1d6434b8b29ae775ad8c2e48c5391)
> @@ -1 +0,0 @@
> -#include <linux/ioctl32.h>
> Index: include/asm-um/elf.h
> ===================================================================
> Index: include/asm-x86_64/apicdef.h
> ===================================================================
> 
> which of course doesn't remove that file at all.
> 
> And I bet that when Linus releases patch-2.6.12-rc5.gz and patch-2.6.12.gz,
> they will have the same construct.  AFAICT, the patch-based people will
> need to download a full new tarball to get rid of this dang file.

Feeding --- include/asm-um/elf.h +++ /dev/null patch to cg-patch would make Cogito kill it. No help for regular patch though, I fear. Perhaps some artificial timestamp could help to the file removal heuristic in GNU patch. Or passing it -E, but that will obviously do the wrong thing to any other zero-sized files.

-- 
				Petr "Pasky" Baudis
Stuff: http://pasky.or.cz/
C++: an octopus made by nailing extra legs onto a dog. -- Steve Taylor

Linus Torvalds· May 17, 2005, 21:44 UTC · re: Petr Baudis · lore

On Tue, 17 May 2005, Petr Baudis wrote:

>
> Perhaps some artificial timestamp could help to the file
> removal heuristic in GNU patch. Or passing it -E, but that will
> obviously do the wrong thing to any other zero-sized files.

-E is always correct for the kernel, since zero-length files aren't really supposed to exist anyway, and "make distclean" has always removed them.

		Linus

Linus Torvalds· May 17, 2005, 21:42 UTC · re: Andrew Morton · lore

On Tue, 17 May 2005, Andrew Morton wrote:

> 
> And I bet that when Linus releases patch-2.6.12-rc5.gz and patch-2.6.12.gz,
> they will have the same construct.  AFAICT, the patch-based people will
> need to download a full new tarball to get rid of this dang file.

Or just run "make distclean" once.

Show 5 quoted lines

> It all wouldn't really matter much, except apparently the mere presence of
> this file breaks the UML build.
> 
> Frazzle.  Paolo, I'm almost wondering if we should change that test to also
> check for a zero-length file.

How many people are affected? The file _is_ gone in the git archives, and in fact I wonder if it was ever there, but I didn't bother to check.

		Linus