this breaks builds of head on stable/9 (Re: svn commit: r257268 - head/gnu/usr.bin/binutils/ld

[Brooks Davis]

On Fri, Nov 01, 2013 at 10:32:54PM +0100, Luigi Rizzo wrote:
> On Fri, Nov 01, 2013 at 03:48:48PM -0500, Brooks Davis wrote:
> > On Fri, Nov 01, 2013 at 07:18:42AM +0100, Luigi Rizzo wrote:
> ...
> > > ok so what happens is the following (trying to build HEAD on 9.2, both amd64)
> > > 
> > > - WITHOUT_CLANG and related options:
> > >     "make toolchain"	fails with the error on krb5
> > >     "make buildworld"	works, and a subsequent
> > > 				"make toolchain" also works
> > > 
> > > - without options (so compiling CLANG)
> > > 
> > >     "make toolchain"	fails with the error below
> > > 
> > >     building shared library libc.so.7
> > >     /usr/bin/ld: this linker was not configured to use sysroots
> > >     cc: error: linker command failed with exit code 1 (use -v to see invocation)
> > >     *** [libc.so.7] Error code 1
> > >     1 error
> > >     *** [all] Error code 2
> > >     1 error
> > > 
> > > 
> > > "make toolchain" on an empty tree used to work until before this commit,
> > > and does work if i revert this single change
> > > 
> > > So does this mean that 'toolchain' now has some prerequisites ?
> > 
> > I think I've found it.  I tried the last one first since it was the most
> > bizzare and about 40 lines up there was fairly obvious error output that
> > told me what was wrong.  The issue is that the :U modifer I used is a
> > bmake feature and that the toolchain target doesn't bootstrap bmake (or
> > perhaps doesn't do it early enough).  I plan to commit a workaround for
> > this particular case, but we probably need to alter the toolchain and
> > kernel-toolchain targets to bootstrap bmake on older systems or we'll
> > run into this again with some future change somewhere else.
> 
> yes, bootstrapping bmake early seems a future-proof way to handle this.

I've committed a workaround for HEAD that I'll MFC soon on the basis
that we in theory have infrastructure to support WITHOUT_BMAKE.
(In practice WITHOUT_BMAKE is broken and corrupts your object tree with a
make that fails until manually deleted.)  We should decide if it should
be fixed or removed before 10 goes out.

-- Brooks
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 188 bytes
Desc: not available
URL: <http://lists.freebsd.org/pipermail/svn-src-all/attachments/20131101/b4b0c7bc/attachment.sig>