On Tuesday 30 December 2008 07:49:36 Christoph Hellwig wrote:
> On Tue, Dec 30, 2008 at 06:52:24AM -0500, Christoph Hellwig wrote:
> > On Mon, Dec 29, 2008 at 04:19:05PM -0600, Eric Sandeen wrote:
> > > This is for http://oss.sgi.com/bugzilla/show_bug.cgi?id=759
> > > "fix parallel issue with include/ symlinks"
> > >
> > > Doing something like "make -j2 install install-dev" would lead
> > > to a race of removing & creating symlinks in include/
> > >
> > > Thanks to Mike for reporting & suggesting the fix.
> > Looks good and I've put in in my queue.
> Actually spoken too soon. This breaks the acl build for me with:
> gcc -g -O2 -DNDEBUG -funsigned-char -fno-strict-aliasing -Wall
> -DVERSION=\"2.2.47\" -DLOCALEDIR=\"/usr/share/locale\"
> -DPACKAGE=\"acl\" -I../include -D_GNU_SOURCE -D_FILE_OFFSET_BITS=64
> -include perm_copy.h -c acl_get_fd.c -fPIC -DPIC -o .libs/acl_get_fd.o
> acl_get_fd.c:27:24: error: acl/libacl.h: No such file or directory
> acl_get_fd.c: In function 'acl_get_fd':
> acl_get_fd.c:62: warning: implicit declaration of function
> acl_get_fd.c:62: warning: return makes pointer from integer without a
> It looks like the acl link in include/ doesn't get created properly
> with this patch.
acl-2.2.47 built fine for me with `./configure && make` and this patch ...
that said, if the only purpose of this is to make sure there are symlinks in
include/, why not do it at the tail end of configure ? then there will be no
weird parallel magic to worry about, and everything will always be available
once configure has finished.
Description: This is a digitally signed message part.