[CMake] 2.8.3 fortran regression or how to find configure problem

[Brad King]

On 11/8/2010 3:54 PM, Allen D Byrne wrote:
>    The projects do differ in the VFFortranCompilerTool option of the project file
> 
> DEBUG:
> 
> AdditionalOptions=" /W1 /libs:dll /threads /dbglibs" instead of
> AdditionalOptions=" /dbglibs"
> 
> RELEASE:
> 
> AdditionalOptions=" /W1 /libs:dll /threads" instead of ""
> 
> and
> 
> DEBUG:
> 
> RELEASE:
> 
> SuppressStartupBanner="true" Preprocess="preprocessYes" instead of ""
> 
> Taking out the new AdditionalOptions=" /W1 /libs:dll /threads" allowed
> the project to succeed.

The regression was introduced here:

  http://cmake.org/gitweb?p=cmake.git;a=commitdiff;h=20f49730

but it was a fix for another bug:

  http://www.cmake.org/pipermail/cmake/2010-September/039829.html

I think this regression was an existing bug that was exposed
by the other fix.  A table in

  Source/cmLocalVisualStudio7Generator.cxx

that maps from compiler flags to IDE project file entries has
no entries for these flags.  They should not be put in the
AdditionalOptions entry but instead mapped to other attributes.

The current behavior probably causes conflicting flags to
be passed by the IDE to the compiler.  I don't know why this
problem didn't show up while testing the fix for the above
bug.  I'll look at adding the missing flag mapping entries.

-Brad