From mboxrd@z Thu Jan 1 00:00:00 1970 From: Mathieu Desnoyers Subject: Re: [PATCH] kselftest: replace $(RM) with rm -f command Date: Sat, 3 Oct 2015 18:05:03 +0000 (UTC) Message-ID: <939391580.17560.1443895503898.JavaMail.zimbra@efficios.com> References: <20150925154415.GC38748@vmdeb7> <1443406217-137773-1-git-send-email-long.wanglong@huawei.com> <1201012824.5792.1443410213729.JavaMail.zimbra@efficios.com> <20151003043808.GA90454@vmdeb7> <1486799898.16862.1443881517178.JavaMail.zimbra@efficios.com> <20151003175528.GA4455@x> Mime-Version: 1.0 Content-Type: text/plain; charset=utf-8 Content-Transfer-Encoding: 7bit Return-path: In-Reply-To: <20151003175528.GA4455@x> Sender: linux-kernel-owner@vger.kernel.org To: Josh Triplett Cc: dvhart , Wang Long , shuahkh@osg.samsung.com, Michael Ellerman , keescook@chromium.org, davem@davemloft.net, luto@amacapital.net, wad@chromium.org, Andrew Morton , gorcunov , cov@codeaurora.org, bobby prani , tyler baker , Tim Bird , Andrea Arcangeli , andrej skvortzov , sjayaram , treding@nvidia.com, Ingo Molnar , naresh kamboju , alexey kodanev , linux-api , linux-kernel@vger.kernel.org, wanglong List-Id: linux-api@vger.kernel.org ----- On Oct 3, 2015, at 1:55 PM, Josh Triplett josh@joshtriplett.org wrote: > On Sat, Oct 03, 2015 at 02:11:57PM +0000, Mathieu Desnoyers wrote: >> ----- On Oct 3, 2015, at 12:38 AM, dvhart dvhart@infradead.org wrote: >> >> > On Mon, Sep 28, 2015 at 03:16:53AM +0000, Mathieu Desnoyers wrote: >> >> ----- On Sep 27, 2015, at 10:10 PM, Wang Long long.wanglong@huawei.com wrote: >> >> >> >> > Some test's Makefile using "$(RM)" while the other's >> >> > using "rm -f". It is better to use one of them in all >> >> > tests. >> >> >> >> I agree that this disparity appears to be unwanted. We >> >> should settle on one or the other. >> >> >> >> > >> >> > "rm -f" is better, because it is less magic, and everyone >> >> > konws what is does. >> >> >> >> "$(RM)" is clearly defined as a Makefile implicit variable >> >> which defaults to "rm -f". >> >> Ref. https://www.gnu.org/software/make/manual/html_node/Implicit-Variables.html >> >> >> >> Leaving it as a variable is more flexible because then the >> >> default behavior can be overridden if need be, which is >> >> not the case of a hardcoded "rm -f". >> >> >> >> Following your line of argumentation, we should then >> >> invoke "gcc" directly in every Makefile because it is >> >> less magic than "$(CC)". This makes no sense. >> > >> > I don't think they can be compared so simply. Specifying a compiler is a common >> > use case. Customizing the rm command is not, in my experience anyway, and like >> > Michael, I would definately have to look up what RM means. >> > >> > That said, I care more about consistency than which is used. Both are valid, but >> > $(RM), while more flexible, will cost more people time to look up what it does >> > as it isn't commonly used than any benefit we're likely to see from its use. >> > >> > Meh. :-) >> >> An example is "grm" when you install the opencsw repository >> packages on Solaris. In the unlikely example where someone >> would have a Solaris machine to build Linux, overriding >> various command names, including "rm", can be useful. This >> is just one example, there are probably others. > > Does Solaris rm not support -f? Yes, it does. I was merely showing this as an example where it can be useful to override the command name, although I don't expect anyone to have to use "grm" rather than "rm" on that specific platform. Thanks, Mathieu -- Mathieu Desnoyers EfficiOS Inc. http://www.efficios.com