Re: [Gcl-devel] How to build gcl? - build worked with -fno-stack-protector

[Tim Josling]

The build worked OK. I'll install it and try it out.

Regards
Tim Josling

On Thu, 2008-01-03 at 18:38 -0500, Camm Maguire wrote:
> I hope the ubuntu package follows Debian.  If so, GCL_PROF=t gclcvs
> should run a profiling image.  GCL_ANSI is the other controlling
> environment variable -- the package has four images in all.  If this
> is not the case, we should followup with whoever disabled it.
> 
> On Debian, the package will setup defaults for these variables to your
> specification as well.
> 

Yep that works though the Ubuntu one is not the latest (all times on the
profile are reported as zero). Encouraging progress though!

> This symbol (__stack_chk_fail) was added by your compiler -- I've
> never seen it before.  You might find the following helpful:
> 
>
http://hackinglinux.blogspot.com/2006/11/resolving-stackchkfail-error.html
> 
> Please keep me posted.
> 

I changed the code to add the -fno-stack-protector and the make is still
underway and past the previous failure point. I will let you know how it
goes.

Here is the diff for the changes I put in - a temporary hack obviously.
Is there a more elegant way to change the compiler options? The (format
*error-output*...) is just to make sure it worked.

        RCS file: /sources/gcl/gcl/cmpnew/gcl_cmpmain.lsp,v
        retrieving revision 1.68
        diff -r1.68 gcl_cmpmain.lsp
        549c549,550
        <    (format nil  "~a -I~a ~a ~a -c ~a -o ~a ~a"
        ---
        >    (let ((ff
        >    (format nil  "~a -fno-stack-protector -I~a ~a ~a -c ~a -o
        ~a ~a" ;; tj
        579c580,582
        <          )
        ---
        >          )))
        >      (format *error-output* "~%tj: compiler command ~S~%" ff)
        >      ff)

By the way I have a 4-cpu system but make -j n (parallel make) fails. Is
this likely to be easy to fix (if so I will probably have a look at
doing that)?

> Tim, I'd appreciate your feedback on the auto-recompilation and source
> inlining features of cvs head when you get a chance.  You might also
> notice significantly enhanced tail call recursion and mutual recursion
> elimination facilities.

Happy to provide feedback ... do I need to turn these options on; if so
how would I do that?

Thanks,
Tim Josling