On Wed, 2012-07-25 12:24:03 +0200, Jordi Guillaumes i Pons 
<[email protected]> wrote:
> I've got a basic idea about which is the problem with ld. It's
> related to LIBPATH. For some reason it puts the contents of LIBPATH
> in the generated script:
> 
> /* Default linker script, for normal executables */
> OUTPUT_FORMAT("a.out-pdp11", "a.out-pdp11",
>               "a.out-pdp11")
> OUTPUT_ARCH(pdp11)
> SEARCH_DIR("/usr/local/pdp11-aout/lib");
> /opt/orahome/tuxedo10gR3/lib:/usr/lib/jvm/java-6-sun/lib/i386/server:/usr/lib/jvm/java-6-sun/jre/bin:/opt/cobol-it/lib:
> PROVIDE (__stack = 0);
> SECTIONS

Oh... Not so funny. (I mean the "orahome" part, not the inclusion of
$LIBPATH.)  Some time ago, I did do OCI hacking. That was totally k3w1
and fun. Not.  I've rarely seen an API interface *that* bad designed
and *that* hard to use.

Looking at the generating script, it works as designed. LIBPATH
shouldn't be set while configure'ing and building binutils and gcc.

Does GCC now work for you, too?

> I understand it should not even try to use the native LIBPATH in a
> cross-build environment. Actually, it looks like some bug in the
> genscripts.sh script, since it just spews $LIBPATH without proper
> "bracketing" (ie, using SEARCH_DIR($LIBPATH)).

Judging from the comments, it works as designed. However, it would be
nice if ./configure or  make  would catch this.

> I've manually edited the generated scripts and the epdp11.c source
> file and I've rebuilt binutils. Now it works OK. I'll try to file a
> bug report about this.

I'm waiting for a success report for GCC from you :)

MfG, JBG

-- 
      Jan-Benedict Glaw      [email protected]              +49-172-7608481
Signature of:            http://www.chiark.greenend.org.uk/~sgtatham/bugs.html
the second  :

Attachment: signature.asc
Description: Digital signature

_______________________________________________
Simh mailing list
[email protected]
http://mailman.trailing-edge.com/mailman/listinfo/simh

Reply via email to