Re: [Gcl-devel] mac os x port questions

[Camm Maguire]

Greetings!

Aurelien Chanudet <address@hidden> writes:

> Hi Camm !
> 
>  > Apart from the time needed to commit, how stable is your bfd work
>  > here?
> 
> I'm meticulously carrying out unit testing for my bfd work. For that
> matter, I wrote a sample test program (much like "bfdtest.c") and as
> exhaustive a test suite as I could, trying to test every possible
> relocation pattern for the Mach-O ABI. As far as I can tell, it's
> fairly stable. ACL2 and Maxima are working fine.
> 

Are your bfd changes in the form of additional files, or do you modify
existing files used by other systems?

Given the stability your report, here's what I think would be best for
2.6.x:  

1) have configure automatically set the appropriate switches.  How
   does config.guess report your machine?  I'm thinking the following
   should be the automatically determined defaults :
   --enable-machine=powerpc-macosx --disable-statsysbfd
   --enable-custreloc
2) I'm not sure how far away sgc is, but it seems to be a matter of a
   few trivial #defines.  Right now, the following need to be set when
   defining SGC:
        MPROTECT_ACTION_FLAGS
        INSTALL_MPROTECT_HANDLER
        GET_FAULT_ADDRESS
        INSTALL_SEGMENTATION_CATCHER

   I only see a definition of the last in your file.  If I'm
   overlooking something simple, we can get it in quickly for 2.6.x,
   otherwise we should add #undef SGC
3) If your bfd stuff is in the form of additional files, as opposed to
   modifications, we can get them in 2.6.2 and let the user choose the
   option --enable-locbfd if they desire.  Otherwise we'll wait for
   2.7.x. 

These are just my thoughts from your reports -- of course you know
best here and I'll defer to your judgement.  I'd just like 2.6.2 to
getnerate as few macosx bug reports as possible :-).

At some point, it would be nice if someone could report on the build
status of 2.6.2 with the latest commits.  The configure options are
not yet automatic, and I think #undef SGC will be required until we
get the other defines in, but I'm thinking those should be the only
issues.  Confirmation appreciated.




>  > Does this build save loaded binary objects so that they are found on
>  > image restart?
> 
> Yes it does, but it's far from being optimal : the saved image is
> dramatically large, comparing to what I get with BFD. In big picture
> terms, this is because the BFD method directly works on raw (.o)
> object files as output by the compiler, whereas this alternate method
> requires transforming these raw object files into larger shared object
> (.so) files. BFD is also much better as far as speed is concerned.
> 
>  > Can I backport your macosx files to the stable branch?  Are they
>  > ready?  These are to my understanding:
> 
> Yes.

Just did this.

Take care,

> 
>  > Time permitting, it would be great to have 2.6.2 work without extra
>  > configure switches for as many of gcl,maxima,acl2, and axiom as
>  > possible.  What was the last test status for these builds?
> 
> Sure. I've never tried axiom, but my last builds for gcl, maxima and
> acl2 successfully pass the tests.
> 
> 
> 
> _______________________________________________
> Gcl-devel mailing list
> address@hidden
> http://mail.gnu.org/mailman/listinfo/gcl-devel
> 
> 
> 

-- 
Camm Maguire                                            address@hidden
==========================================================================
"The earth is but one country, and mankind its citizens."  --  Baha'u'llah