Hi Werner: I changed the subject line to something more appropriate since this thread may continue for a while.
On 2009-08-14 08:55+0200 Werner Smekal wrote: > Hi Alan, >> >> In addition after I configure PLplot with cmake and run make, cmake is run >> again due "Re-run cmake: build system dependency is missing" (see >> make.out). It might have something to do with a misconfigured bindings, >> since when I turn all bindings off, this doesn't happen. I think the re-run >> should also not happen. I try to find out, which bindings causes this. > > I found out what causes the rerun. I have to disable f77 and f95 bindings at > the command line, then a make call doesn't lead to an rerun of cmake. I > attached both outputs of cmake. The command line was > > cmake -DBUILD_TEST=ON ../plplot &> cmake.out &> cmake.out > cmake -DBUILD_TEST=ON -DENABLE_f77=OFF -DENABLE_f95=OFF ../plplot &> > cmake_wo_fortran.out > > The difference between these files is > > < -- The Fortran compiler identification is unknown > < -- Configuring done > < -- Generating done > < -- Build files have been written to: > /Users/smekal/Development/plbuild/language_tests/Fortran > < -- The Fortran compiler identification is unknown > < -- WARNING: no working Fortran compiler so disabling Fortran bindings and > examples. > < -- CMAKE_GENERATOR = Unix Makefiles > > Is this behavior reproducible on Linux (where no working Fortran compiler is > available)? > > Let me know if you need further information. Valery Pipin has also reported the cmake re-run issue (on AltLinux, I believe). I had a detailed look this morning with cmake-2.6.4, and I could not reproduce the issue at all by simply disabling gfortran. However, by attempting to get as close as possible to what you did, I finally found a case where cmake reran on the first "make" attempt (but not subsequent ones for some reason). At this point I have no idea which of my seemingly innocuous steps to reproduce (for example, using "make" rather than my usual "make -j4 install", dropping the FC environment variable, and disabling gnatmake as well as gfortran) was the essential one that triggered the cmake re-run. I haven't tried it myself yet, but another thing that surprises me is your setting of ENABLE_f77 and ENABLE_f95 to OFF made a difference for you. After all, your cmake.out and mine confirm that the first cmake run does this automatically by itself when it cannot find a fortran compiler. So I suspect you got a different result because of some slight other difference in how you invoked your test. There are still a lot of questions here, and the most innocuous details seem to matter so I think the best thing to do is to try to find the simplest case where I can reliably generate a cmake rerun, then start debugging with liberal use of the message command. Normally, I understand why cmake would be re-run when a compiler was missing since it does everything in its power to find compilers assuming that the user will immediately install one when he encounters the first missing compiler message. But our build system ignores all that (except for C and C++ which are special cases) and bypasses the compiler search altogether for subsequent runs (e.g., by setting both ENABLE_f77 and ENABLE_f95 to OFF) if the first attempt to find a compiler fails. Anyhow, I will keep investigating since there is no way that a cmake re-run should be triggered by a missing compiler when we supress all subsequent searches for that compiler. Also, I encourage you to continue your own independent investigation (being careful about any slight variations in the way you do your tests) since you might well find a solution to the problem before I do. Alan __________________________ Alan W. Irwin Astronomical research affiliation with Department of Physics and Astronomy, University of Victoria (astrowww.phys.uvic.ca). Programming affiliations with the FreeEOS equation-of-state implementation for stellar interiors (freeeos.sf.net); PLplot scientific plotting software package (plplot.org); the libLASi project (unifont.org/lasi); the Loads of Linux Links project (loll.sf.net); and the Linux Brochure Project (lbproject.sf.net). __________________________ Linux-powered Science __________________________ ------------------------------------------------------------------------------ Let Crystal Reports handle the reporting - Free Crystal Reports 2008 30-Day trial. Simplify your report design, integration and deployment - and focus on what you do best, core application coding. Discover what's new with Crystal Reports now. http://p.sf.net/sfu/bobj-july _______________________________________________ Plplot-devel mailing list Plplot-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/plplot-devel
