Re: [Mingw-users] gcc 4.1.1 include search order

[Keith M. <kei...@to...>]

Michael Gerdau wrote, in dialogue with Danny Smith:
>>> I haven't looked into it for some time now but IIRC the cause for
>>> the xcompiler was related to [not] properly (or so) detecting the
>>> usr-local include path of the target environment.
>> 
>> 
>> And that problem can be fixed by using the --with-sysroot=<DIR>
>> configure opt. See gccinstall.info section on Cross-Compiler Specific
>> Options and the --sysroot compile option in gcc.info.
>
> That's what I and a few others thought it was. Unfortunately it
> does not have the desired effect :-/
> (details of he discussion is in the archives; couple of months ago)
>
> Whether this is due to a bug in the configure scripts or anything
> else I don't know.

Hmm.  This looked the most promising avenue of investigation; shame
that it doesn't seem to do what it says on the tin.  Have you tried
`--with-headers=<DIR>' instead?  The `info' file I'm reading says
that's been deprecated in favour of `--with-sysroot', but perhaps
this isn't the case for 3.4.x; (I haven't had time to research this
properly yet, sorry).

> However last night I have created a crosscompiler incorporating
> the Patch from Mark Mitchell as of
>   http://gcc.gnu.org/ml/gcc-patches/2005-08/msg00122.html
>
> Later today I will check whether my newly build gcc now does things
> as desired.

We can live in hope, but I'm not optimistic.  The docs relating to
that patch say that the `NATIVE_SYSTEM_HEADER_DIR' macro, which it
defines, is associated with `SYSTEM_INCLUDE_DIR' preprocessor macro,
of which the `info' file says `Cross compilers do not use this macro,
and do not search the directory specified'.

Nonetheless, I too will likely adopt Mark Mitchell's patch -- it will
give me a good opportunity to test the `autopatch' feature I've added
to my modular, and interactive, variation of your build script, Michael,
which I hope to be ready to release into the wild within the next week
or two.

Regards,
Keith.