[Bug fortran/58276] "make install-host" gfortran is not functional

["baker at usgs dot gov" <gcc-bugzilla at gcc dot gnu dot org>]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58276

--- Comment #7 from Larry Baker <baker at usgs dot gov> ---
Andrew,

On 29 Aug 2013, at 4:50 PM, pinskia at gcc dot gnu.org wrote:

>> Where can I read about the distinction between "make install", "make
>> install-host", and "make install-target"?  Is "make install-host" supposed
>> to install usable compilers?
> 
> Define usable compiler?  A compiler which can be used to compile the target
> libraries (including glibc)?  Then yes install-host is enough for that.  If you
> define it as usable for full cross compiling, then no it is not enough.

I know about the need to build a gcc twice for a cross development tool chain:
once to build the minimal compiler to build the target libraries, and the final
build for the tool chain.  bitbake generates full cross development tool
chains.  From your description, "make install-host" is not sufficient; "make
install" would be.  When I have made my own gfortran ARM cross compiler, I have
always done a "make install", so it has been functional.

I will investigate, but I think it is likely that the bitbake build system does
the library installs in a separate recipe, and gcc and g++ support libraries
are enabled by default.  That would explain why their gcc and g++ cross
compilers are functional and gfortran is not.  I did not request a separate
library installation step for libgfortran, since I never had to do that with my
own build.  Now that I know there are separate make installs, I think I can
make progress.

Thank you for your help.

Larry Baker
US Geological Survey
650-329-5608
baker@usgs.gov