Re: blcr 0.8.2 with ubuntu karmic (kernel 2.6.31)

From: Paul H. Hargrove (PHHargrove_at_lbl_dot_gov)
Date: Tue Nov 10 2009 - 15:52:13 PST

  • Next message: Chris Chaudoir: "Re: Memory Exclusion"
    Alan Woodland wrote:
    > 2009/11/10 Paul H. Hargrove <PHHargrove_at_lbl_dot_gov>:
    >   
    >> Alan Woodland wrote:
    >>     
    > [snip]
    >   
    >>> The tests passed, except prctl(). The problem with the prctl test is
    >>> actually that the configure test lives inside the "if cr_build_libcr"
    >>> conditional, but I've been using --with-installed-libcr because the
    >>> tests are the only part of the package that doesn't end up in a Debian
    >>> package. If I don't configure with this flag it passes. I guess the
    >>> fix is to either always run this test in configure, or do it when
    >>> building one of the tests or the library?
    >>>
    >>> Alan
    >>>
    >>>       
    >> OK, so I thing I messed up putting the HAVEW_PRCTL configure probe
    >> conditional on cr_build_libcr, when it should probably have been conditional
    >> on cr_build_tests.  The attached patch should resolve that.
    >>     
    > Haven't had a chance to test the patch yet.
    >
    >   
    >> However, IIRC this test should be printing "SKIPPED" rather than "FAIL" if
    >> the configure probe failed (or in this case it was not run).  Is that right?
    >>     
    > The actual error was:
    >
    > No prctl() support was found at configure time - test skipped
    > checkpoint/nonzeroexit (77)
    > FAIL: prctl.ct
    >
    > Other tests print skipped as expected (e.g. hugetlbfs, bug2524). The
    > prctl test seems to be hitting:
    > #ifndef HAVE_PRCTL
    > int main(void) {
    >     printf("No prctl() support was found at configure time - test skipped\n");
    >     return 77;
    > }
    > #else
    > from prctl.c, so it looks like it's returning the magic test skipped
    > value and I can't quite see a scenario that would lead to the test
    > caller not noticing it right now.
    >
    > Alan
    >   
    
    The prctl test is run by the "crut_wrapper", not directly from the 
    Makefile.  So, I am guessing that this is an oversight in crut_wrapper.c 
    (which I don't plan to fix immediately).  If the configure patch 
    resolves the problem then I think we can stick with that until the next 
    full release.
    
    -Paul
    
    -- 
    Paul H. Hargrove                          PHHargrove_at_lbl_dot_gov
    Future Technologies Group                 Tel: +1-510-495-2352
    HPC Research Department                   Fax: +1-510-486-6900
    Lawrence Berkeley National Laboratory     
    

  • Next message: Chris Chaudoir: "Re: Memory Exclusion"