From mboxrd@z Thu Jan 1 00:00:00 1970 X-Spam-Checker-Version: SpamAssassin 3.4.4 (2020-01-24) on inbox.vuxu.org X-Spam-Level: X-Spam-Status: No, score=-2.0 required=5.0 tests=MAILING_LIST_MULTI, RCVD_IN_DNSWL_MED,RCVD_IN_MSPIKE_H3,RCVD_IN_MSPIKE_WL, SUBJ_OBFU_PUNCT_FEW,SUBJ_OBFU_PUNCT_MANY autolearn=ham autolearn_force=no version=3.4.4 Received: (qmail 4753 invoked from network); 24 Aug 2020 16:17:05 -0000 Received: from mother.openwall.net (195.42.179.200) by inbox.vuxu.org with ESMTPUTF8; 24 Aug 2020 16:17:05 -0000 Received: (qmail 4035 invoked by uid 550); 24 Aug 2020 16:16:59 -0000 Mailing-List: contact musl-help@lists.openwall.com; run by ezmlm Precedence: bulk List-Post: List-Help: List-Unsubscribe: List-Subscribe: List-ID: Reply-To: musl@lists.openwall.com Received: (qmail 4017 invoked from network); 24 Aug 2020 16:16:58 -0000 Date: Mon, 24 Aug 2020 12:16:47 -0400 From: Rich Felker To: musl@lists.openwall.com Message-ID: <20200824161646.GH3265@brightrain.aerifal.cx> References: <5232EC15-7E60-4D1F-BF3B-C31BFF998C06@rb67.eu> MIME-Version: 1.0 Content-Type: text/plain; charset=utf-8 Content-Disposition: inline Content-Transfer-Encoding: 8bit In-Reply-To: <5232EC15-7E60-4D1F-BF3B-C31BFF998C06@rb67.eu> User-Agent: Mutt/1.5.21 (2010-09-15) Subject: Re: [musl] Incompatible behaviour of res_query(3) w.r.t. NXDOMAIN On Sun, Aug 23, 2020 at 11:31:32PM +0200, Daniel Neri wrote: > Hello, > > > Musl’s res_query(3) returns success for a query that results in an > NXDOMAIN response, which disagrees with other common > implementations. > > Simple test case: https://t.rb67.eu/res_query.c > > On Debian Linux, OpenBSD and FreeBSD this program prints: > > error: Unknown host > > but on Alpine Linux edge (musl v1.2.1): > > success (size 91) I think there's a good argument that this should be changed, but that would require making res_* set h_errno, which would require making h_errno thread-local -- which it probably should have been all along, but the standard interfaces that used it, gethostby* etc., were non-thread-safe anyway and thus it seemed unnecessary. FWIW I think the musl behavior is *better* here (more informative to the application), but usually it's preferable not to break the existing (even if informal) contracts of the interfaces being implemented than to be "better". Rich