From mboxrd@z Thu Jan 1 00:00:00 1970 From: Baruch Siach Date: Thu, 22 Oct 2015 20:00:55 +0300 Subject: [Buildroot] ltp-testsuite without native RPC (commit 8f1c4c49a2ab0fc9e6cf73370f4483e253e5ac2d) In-Reply-To: References: <20151015181727.GI2451@tarshish> <20151019173523.GX20889@tarshish> <20151020034901.GZ20889@tarshish> Message-ID: <20151022170055.GE5512@tarshish> List-Id: MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit To: buildroot@busybox.net Hi Thomas, On Thu, Oct 22, 2015 at 10:24:24AM +0200, Thomas De Schampheleire wrote: > On Tue, Oct 20, 2015 at 5:49 AM, Baruch Siach wrote: > > On Mon, Oct 19, 2015 at 08:35:23PM +0300, Baruch Siach wrote: > >> On Mon, Oct 19, 2015 at 03:26:18PM +0200, Thomas De Schampheleire wrote: > >> > On Thu, Oct 15, 2015 at 8:17 PM, Baruch Siach wrote: > >> > > On Thu, Oct 15, 2015 at 04:06:39PM +0200, Thomas De Schampheleire wrote: > >> > >> What about patching ltp-testsuite instead, not enabling the RPC tests > >> > >> in this case? At first sight this was added with following (ltp) > >> > >> commit: > >> > >> https://github.com/linux-test-project/ltp/commit/8e9a4dbb8eaa561e51d6d8ebe5a5342f3e498224 > >> > > > >> > > Isn't that what ac_cv_header_tirpc_netconfig_h=no is doing? > >> > > >> > I just tried building ltp-testsuite with a no-rpc toolchain on top of > >> > latest buildroot. Here is the build failure: > >> > > >> > /usr/bin/make -C lib -f > >> > ".../output/build/ltp-testsuite-20150420/testcases/network/rpc/basic_tests/rpc01/lib/Makefile" > >> > all > >> > make[7]: Entering directory > >> > `.../output/build/ltp-testsuite-20150420/testcases/network/rpc/basic_tests/rpc01/lib' > >> > .../output/host/usr/bin/powerpc-linux-gcc -D_LARGEFILE_SOURCE > >> > -D_LARGEFILE64_SOURCE -Os -g -O2 -fno-strict-aliasing -pipe -Wall -W > >> > -D_LARGEFILE_SOURCE -D_LARGEFILE64_SOURCE -Os -Wold-style-definition > >> > -D_LARGEFILE_SOURCE -D_LARGEFILE64_SOURCE -D_FORTIFY_SOURCE=2 > >> > -I../../../../../../include -I../../../../../../include -c -o > >> > librpc01.o librpc01.c > >> > librpc01.c:19:21: fatal error: rpc/xdr.h: No such file or directory > >> > #include > >> > ^ > >> > compilation terminated. > >> > make[7]: *** [librpc01.o] Error 1 > >> > > >> > > >> > I think my original problem (this was with a previous version of > >> > ltp-testsuite) was that rpc/rpc.h was missing, but I think the bottom > >> > line is the same: ltp-testsuite attempts to build RPC-related tests > >> > but the necessary header files are not available. > >> > > >> > My approach was to add libtirpc to the equation, which provides the > >> > header files and library to make things work. > >> > > >> > With your patch, libtirpc is not added, and so the header files remain > >> > unavailable. > >> > > >> > How do you suggest to proceed? > >> > One approach is to disable the ltp tests that use rpc entirely (on top > >> > of your last patch), another one is to disable only the tests that use > >> > authdes_create (on top of my original patch; reverting yours). > >> > Other ideas are welcome, of course. > >> > >> The best solution seems to be fixing the root cause upstream. There was an > >> attempt by Natanael Copa in commit c0547c56dafb8e1 (Add configure option to > >> disable DES authentication) to add a --disable-authdes configure option to > >> libtirpc. There was another attempt in commit 58cd17f999 (Fix: unresolved > >> _crypt() symbols) to add a local implementation of DES. The first attempt was > >> later reverted in commit f17b44048003 (Revert commit c0547c56dafb ('Add > >> configure option to disable DES')) which also (inadvertedly?) removed > >> des_crypt.c and des_impl.c from Makefile.am. Restoring the local DES > >> implementation looks not so hard to do. > >> > >> In the mean time I think we can replace our > >> 0007-Disable-DES-authentification-support.patch with something more flexible > >> along the lines of Natanael's reverted patch. > > > > Which, of course, won't solve our problem. > > > > I'll look into the first approach (libtirpc local DES) if I find some time. > > Thanks in advance, Baruch! It turns out that making auth_des to build takes more than a DES implementation. NIS support is also needed. At this point I think your original suggestion, i.e. disabling authdes_create tests, makes more sense. I'm not sure when I'll be able to look into this again, though. baruch -- http://baruch.siach.name/blog/ ~. .~ Tk Open Systems =}------------------------------------------------ooO--U--Ooo------------{= - baruch at tkos.co.il - tel: +972.2.679.5364, http://www.tkos.co.il -