From mboxrd@z Thu Jan 1 00:00:00 1970 From: SZEDER =?iso-8859-1?Q?G=E1bor?= Subject: Re: [PATCH 2/2] bash: not all 'git bisect' subcommands make sense when not bisecting Date: Mon, 11 Oct 2010 00:12:32 +0200 Message-ID: <20101010221232.GD12529@neumann> References: <1286746774-7877-1-git-send-email-szeder@ira.uka.de> <1286746774-7877-2-git-send-email-szeder@ira.uka.de> Mime-Version: 1.0 Content-Type: text/plain; charset=iso-8859-1 Content-Transfer-Encoding: QUOTED-PRINTABLE Cc: "Shawn O. Pearce" , Junio C Hamano , git@vger.kernel.org To: Sverre Rabbelier X-From: git-owner@vger.kernel.org Mon Oct 11 00:13:33 2010 Return-path: Envelope-to: gcvg-git-2@lo.gmane.org Received: from vger.kernel.org ([209.132.180.67]) by lo.gmane.org with esmtp (Exim 4.69) (envelope-from ) id 1P5491-0008QY-3l for gcvg-git-2@lo.gmane.org; Mon, 11 Oct 2010 00:13:27 +0200 Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1752083Ab0JJWMk convert rfc822-to-quoted-printable (ORCPT ); Sun, 10 Oct 2010 18:12:40 -0400 Received: from moutng.kundenserver.de ([212.227.17.9]:59105 "EHLO moutng.kundenserver.de" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1752131Ab0JJWMk (ORCPT ); Sun, 10 Oct 2010 18:12:40 -0400 Received: from [127.0.1.1] (p5B130D54.dip0.t-ipconnect.de [91.19.13.84]) by mrelayeu.kundenserver.de (node=mreu1) with ESMTP (Nemesis) id 0MAysS-1OxKuD2YB2-00AX4N; Mon, 11 Oct 2010 00:12:34 +0200 Content-Disposition: inline In-Reply-To: User-Agent: Mutt/1.5.18 (2008-05-17) X-Provags-ID: V02:K0:X5IJt41W8YWUI+RcKFQ0o1PpMWIKLv89QbSuWZYBLWy ZRyRDRCT/hzdh7O61pEBF1myy64vc/HCiSpMWwZLPNWpBGDYYm 7hFjraBCc4vizKHRqoxYATpPEt1kQoI5A+cjsF8J28R/qGFq2d ZS3CK/885vuaT91kqtVmu7EHgH3aUXujo8xFh+Zci432/SH/JM 0NQphmSd8sgZZnJA/jBsw== Sender: git-owner@vger.kernel.org Precedence: bulk List-ID: X-Mailing-List: git@vger.kernel.org Archived-At: Hi, On Mon, Oct 11, 2010 at 12:05:52AM +0200, Sverre Rabbelier wrote: > Heya, >=20 > On Sun, Oct 10, 2010 at 23:39, SZEDER G=E1bor wro= te: > > ... but only 'start' and 'replay'. =A0The other commands will eithe= r > > error out or offer to start bisecting for the user. >=20 > I'm not sure, following the same concept "git rebase" should not offe= r > "--continue" if you're not rebasing. Actually, my patch follows the same concept of 'git rebase', because that's exactly what the completion function for 'git rebase' does: it doesn't offer '--continue' (nor '--abort' and '--skip') if you're not rebasing. Best, G=E1bor