From mboxrd@z Thu Jan 1 00:00:00 1970 From: Ian Jackson Subject: Re: [PATCH] hg ignore libxl lex/yacc detritus Date: Fri, 23 Jul 2010 16:53:49 +0100 Message-ID: <19529.47885.676254.574852@mariner.uk.xensource.com> References: Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit Return-path: In-Reply-To: List-Unsubscribe: , List-Post: List-Help: List-Subscribe: , Sender: xen-devel-bounces@lists.xensource.com Errors-To: xen-devel-bounces@lists.xensource.com To: Keir Fraser Cc: Dan Magenheimer , "xen-devel@lists.xensource.com" , Stabellini Stefano List-Id: xen-devel@lists.xenproject.org Keir Fraser writes ("Re: [Xen-devel] [PATCH] hg ignore libxl lex/yacc detritus"): > My preference would be to delete the checked-in generated files. Then either > declare old flex/bison to be simply unsupported, or avoid depending on > whatever feature it is that is not supported in those old flex/bison > versions. Checking in the generated output is a cheap hack workaround. I think what we have now is a good compromise. If you have working flex and bison then everything works absolutely fine; if we update the input files we have to remember to run make before committing but that's not hard (and I do a build test anyway). However providing the checked-in files means that people who have crazy prehistoric versions of flex (1993 in the case of RHEL5 AIUI) or whatever can still build the system. If they work in xen-unstable they may need to knock make on the head, but at least they have a workaround. If we didn't provide the generated files checked-in, the workaround would be much worse. As an aside, I think there is generally nothing wrong in principle with checking in generated files. It's something that should be done with proper thought and care, but it's often appropriate. Eg, automake-using systems should contain configure and configure.in and Makefile.in, so that people who don't have the right version of automake can still build from a checkout. Ian.