 |
 |
Hi, Fabrizio On Thursday 31 May 2007 10:51, Fabrizio Furano wrote: > I was just checking the compilation on shire01, and everything was > fine, except for the missing XrdPss. > > Personally, I don't consider one Solaris machine more important as > another as a "reference build system". I just try to compile in more > than one from time to time. Sure. But I think it would be nice to have a number of reference systems on which every release candidate gets tested before a release, so you really know that it should run there. Solaris compilers went through a lot of changes in how they treat templated code, so it can easily happen that you fail with an older one. In autotools the best thing is to have a source directory on a shared filesystem accessible by the machines with the reference architectures. Then for every architecture to test you issue a "make distcheck". This will - generate a distribution tarball - unpack the tarball in a sandbox and build it - test the installation process - make sure that "make DESTDIR=" works. This is important for package builders - run the test suite (I am just writing a few tests) - make sure that "make uninstall" leaves no files (but this is less important, since almost never used) I use this for Linux and Solaris builds at SLAC, because machines of both architectures have access to the afs there. With the right setup and access to machines at CERN one could even automate this further. Cheers, Derek > > Fabrizio > > Derek Feichtinger wrote: > > Hi, Andy > > > > thanks for the fix. But I think a new directory is missing: > > > > The config/GNUmakefile.in holds some lines about a XrdPss target: > > 137a138,142 > > XrdPss: FORCE > > @echo $(XMSG) oss $(XDBG) component... > > $(ECHO)cd src/XrdPss;\ > > $(MAKE) $(MAKEXEQ) ARCH=$(ARCH) --no-print-directory > > > > Could you check these in? > > > > A further question: Is my choice of shire05.slac.stanford.edu as a > > reference build system ok, or should I use another host at slac? What > > compiler environment should I use? > > > > I think I can make an improved build structure for the autobuild. Also I > > have begun writing the beginning of some test scripts, which will make > > some basic functionality tests upon "make check" (start up a daemon, copy > > files, then try more difficult configurations where the modules get > > loaded). This may be a help to the maintainer/distributor, since in > > autotools you can use the special target "make distcheck", which will > > generate a distribution tarball, test building in a sandbox area, test > > compliance with DESTDIR for package generators like RPM, and finally run > > the tests. > > > > Cheers, > > Derek > > > > On Wednesday 30 May 2007 23:32, Andrew Hanushevsky wrote: > >> I have a patch for the problem below. I needed to fix it as well. I > >> wiull commit momentarily. > >> > >> Andy > >> > >>> The newer compiler (2) has a different problem > >>> > >>> Compiling XrdClientMessage.cc > >>> "XrdClientMessage.cc", line 87: Error: The function "posix_memalign" > >>> must have > >>> a prototype. > >>> 1 Error(s) detected. > >>> gmake[4]: *** [../../obj/sun4x_58/XrdClientMessage.o] Error 1 > >>> gmake[3]: *** [SunOSall] Error 2 > >>> gmake[2]: *** [all] Error 2 > >>> gmake[1]: *** [XrdClient] Error 2 > >>> gmake: *** [all] Error 2 > >>> > >>> > >>> Cheers, > >>> Derek > >>> > >>> On Wednesday 30 May 2007 09:54, Gerri Ganis wrote: > >>>> Hi Derek, > >>>> > >>>> Sorry again, I am validating at the CAF a set of changes while > >>>> committing bug fixes > >>>> ... now it should be ok (I've tried the head). > >>>> > >>>> Gerri > >>>> > >>>> Derek Feichtinger wrote: > >>>>> Hi, Gerri > >>>>> > >>>>> Thanks for the fix. But there is yet another problem with a missing > >>>>> function: > >>>>> > >>>>> Compiling XrdClientPhyConnection.cc > >>>>> XrdClientPhyConnection.cc: In function `void* > >>>>> SocketReaderThread(void*, XrdClientThread*)': > >>>>> XrdClientPhyConnection.cc:42: no matching function for call to > >>>>> `XrdClientThread > >>>>> > >>>>> ::MaskSignal(int)' > >>>>> > >>>>> make[4]: *** [../../obj/i386_rhel30/XrdClientPhyConnection.o] Error 1 > >>>>> make[3]: *** [Linuxall] Error 2 > >>>>> make[2]: *** [all] Error 2 > >>>>> make[1]: *** [XrdClient] Error 2 > >>>>> make: *** [all] Error 2 > >>>>> > >>>>> Cheers, > >>>>> Derek > >>>>> > >>>>> On Wednesday 30 May 2007 01:51, Gerri Ganis wrote: > >>>>>> Hi Derek, > >>>>>> > >>>>>> Sorry, I forgot to commit the changes in XrdClient/XrdClientPSock.hh > >>>>>> . Now it should be fixed. > >>>>>> > >>>>>> Cheers, Gerri > >>>>>> > >>>>>> Derek Feichtinger wrote: > >>>>>>> Hi, > >>>>>>> > >>>>>>> there also currently seems to be a problem on Linux with the > >>>>>>> configure.classic build (and surely also the autotools build): > >>>>>>> > >>>>>>> Compiling XrdClientAbs.cc > >>>>>>> In file included from ../XrdClient/XrdClientPhyConnection.hh:18, > >>>>>>> from ../XrdClient/XrdClientLogConnection.hh:21, > >>>>>>> from ../XrdClient/XrdClientConnMgr.hh:29, > >>>>>>> from ../XrdClient/XrdClientConn.hh:23, > >>>>>>> from ../XrdClient/XrdClientAbs.hh:20, > >>>>>>> from XrdClientAbs.cc:16: > >>>>>>> ../XrdClient/XrdClientPSock.hh: In member function `virtual int > >>>>>>> XrdClientPSock::SaveSocket()': > >>>>>>> ../XrdClient/XrdClientPSock.hh:78: `fInterrupt' undeclared (first > >>>>>>> use this function) > >>>>>>> > >>>>>>> Cheers, > >>>>>>> Derek > >>> > >>> -- > >>> Dr. Derek Feichtinger Tel: +41 56 310 47 33 > >>> AIT Group email: [log in to unmask] > >>> PSI > >>> CH-5232 Villigen PSI -- Dr. Derek Feichtinger Tel: +41 56 310 47 33 AIT Group email: [log in to unmask] PSI CH-5232 Villigen PSI