--rqzD5py0kzyFAOWN
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
Content-Transfer-Encoding: quoted-printable

Todd Vierling <tv@duh.org> wrote:
> On Mon, 11 Apr 2005, Jan Schaumann wrote:
>=20
> > In rev 1.49 of  pkgsrc/mk/bulk/post-build, you committed
>=20
> Sorry I haven't gotten back to you on this yet....
>=20
> > -chdir($USR_PKGSRC);
> > -system("find . -name $BROKENFILE -print -o -name $BROKENWRKLOG -print =
| $GTAR -T - -plcf - | (cd $FTP; $GTAR -plxf -)");
> > +chdir($vars{USR_PKGSRC});
> > +my_system("/bin/ls -1 $vars{BROKENFILE} $vars{BROKENWRKLOG} */*/$vars{=
BROKENFILE} */*/$vars{BROKENWRKLOG} 2>/dev/null | $vars{GTAR} -T - -plcf - =
| (cd $vars{FTP}; $vars{GTAR} -plxf -)");
> >
> > On some systems, this now leads to too many arguments to ls(1),
> > subsequently producing an empty bulk-log.  What was wrong with using
> > find(1)?
>=20
> NFS slowness, for a nfs-mounted pkgsrc.
>=20
> Hm.  This is ugly.  I suppose it can go back to find, but I may try to fi=
nd
> a new way to do this in the future.

This still hasn't been fixed.  I'd really like to have a functional
post-build on the new branch -- can we please revert this change to use
find(1) again?  It's better to have it be slow over NFS than to not have
it working at all on some platforms.

-Jan

--=20
42

--rqzD5py0kzyFAOWN
Content-Type: application/pgp-signature
Content-Disposition: inline

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.3 (NetBSD)

iD8DBQFCreBUfFtkr68iakwRAp8vAJ97c1VdZtmXGlVtebhpMSapkqgAYwCgnOGU
lw1NBkRA6Zs3yCU+/h0ByNk=
=qbD6
-----END PGP SIGNATURE-----

--rqzD5py0kzyFAOWN--