On Friday 02 November 2007 18:29:23 James Keenan via RT wrote: > c: Could I ask you to post the error messages you got when this test > failed? Before submitting/committing patches, I ran this test file many > times in its original version on both Darwin and Linux and it never > failed for me.
You probably have libgdbm installed. I think the patch you applied from Cosimo in r22668 may have cleared the error for me. I don't remember the exact test of the message, but the output file from the linking stage said "error with -lgdbm", which means that gcc couldn't link against libgdbm, as I don't have it installed. > Nor would I have expected it to fail. Since it executes > auto::msvc::runstep(), it does exactly what happens every time you say > 'perl Configure.pl'. The fact that runstep() returns true -- which is > what the next-to-last test in the file tests for -- only indicates that > the process of determining "if your C compiler is actually Visual C++" > (as $description reads) completed successfully or not. If you're not on > Windows, the step's 'result' value should be 'undef', as should the > 'msvcversion' element in the Parrot::Configure object -- but runstep() > itself should return true and 113-auto_msvc-01.t should pass all its tests. That sounds reasonable, but on platforms where we know MSVC isn't available, I'm not sure running the tests will give us any interesting or useful information. I'm happy to let you make the final decision as to its value, as you're the expert on this code, but I do think it's worth considering if we get any value out of this test on non-Windows platforms. -- c
