Re: [Devel] freetype-2.1.4: build feedback

[Werner LEMBERG]

> The SGI IRIX native compilers are excellent at diagnosing crud in
> code.  Here is a report (I've wrapped long lines for readability):

Thanks, I've fixed almost everything in the CVS.

> With the SGI C++ and Sun Solaris C++ compilers, there was this
> failure:

Fixed also.

> On GNU/Linux on Alpha, PowerPC, and SPARC, I got link-time failures
> like this: [...]
> 
> The bad linker argument sequence of
> 
>       -rpath -soname libfreetype.so.6
> 
> is the cause of the failure: -rpath requires a directory following
> it, so -soname is gobbled up as that directory, and the linker then
> tries to load libfreetype.so.6, which does not exist yet.
> 
> On GNU/Linux Red Hat 7.2 on Intel IA-32, linking succeeded, but when
> I did "make install", it relinked as above, and died for the same
> reason.

I assume problem with libtool.  Right now I've tried to update to
libtool 1.5 (with `libtoolize --force --copy; aclocal -I .'), but this
seems to be buggy: I get a lot of

  libtool: test: : integer expression expected

warnings; finally, the linker aborts with

  /usr/i486-suse-linux/bin/ld -r
    -o freetype2/objs/.libs//freetype2/objs/libfreetype.la-1.o 
    freetype2/objs/.libs/ftsystem.o 
  /usr/i486-suse-linux/bin/ld: cannot open output file
   freetype2/objs/.libs//freetype2/objs/libfreetype.la-1.o:
     No such file or directory

> Finally, it does not appear that a setting of the variable `prefix'
> in the environment is correctly propagated to the makefiles, as it
> should be according to GNU standards for autoconfiguration.  On
> systems where I lack write access to /usr/local, I had to manually
> set prefix like this:
> 
>       make prefix=$HOME/`arch`/local install

I've checked autoconf.info but I couldn't find any details on an
environment variable called `prefix'.  What exactly do you mean?
Where is it documented?


    Werner