List Info

Thread: intel mac build error
intel mac build error
user name
 2006-02-14 11:15:22 

  


  

    
Hi all,



Tried building io on an intel mac this morning and got:



...

cc -Os -g -Wall -I. -Ibase -ISkipDB -DSANE_POPEN -falign-loops=16 -L.

-L_libs -Lbase/_libs -o io main.c -lIoBase -lIoVM  -ldl

cd ..

./vm/io _build.io

Io Scheduler error: no Coro_setupJmpbuf entry for this platform

.make: *** [all] Error 1





Any ideas?



Best,



-ryan




    




  

  
  
   
     
    

  

    intel mac build error
  


  

     user name

     2006-02-14 15:40:51
  


  

    

  


  

    
On 2/14/06, Ryan Raaum <ryan.raaumgmail.com&gt; wrote:

&gt; Hi all,

>

> Tried building io on an intel mac this morning and got:

>

> ...

> cc -Os -g -Wall -I. -Ibase -ISkipDB -DSANE_POPEN -falign-loops=16 -L.

> -L_libs -Lbase/_libs -o io main.c -lIoBase -lIoVM&nbsp; -ldl

> cd ..

> ./vm/io _build.io

> Io Scheduler error: no Coro_setupJmpbuf entry for this platform

> .make: *** [all] Error 1

>

&gt;

> Any ideas?



Io's coroutine implementation relies on knowledge of how each

platform's stacks work. What will be needed is some investigation on

what the new intel macs use. Then one can edit the correct files so

the build can verify that you are on a supported platform and thus

compile the correct code.



Brian.




    




  

  
  
   
     
    

  

    intel mac build error
  


  

     user name

     2006-02-14 16:21:15
  


  

    

  


  

    
On 2/14/06, Brian Mitchell <binary42gmail.com&gt; wrote:

&gt; On 2/14/06, Ryan Raaum <ryan.raaumgmail.com&gt; wrote:

&gt; > Hi all,

> >

> > Tried building io on an intel mac this morning and got:

> >

> > ...

> > cc -Os -g -Wall -I. -Ibase -ISkipDB -DSANE_POPEN -falign-loops=16 -L.

> > -L_libs -Lbase/_libs -o io main.c -lIoBase -lIoVM&nbsp; -ldl

> > cd ..

> > ./vm/io _build.io

> > Io Scheduler error: no Coro_setupJmpbuf entry for this platform

> > .make: *** [all] Error 1

> >

> >

> > Any ideas?

&gt;

> Io's coroutine implementation relies on knowledge of how each

> platform's stacks work. What will be needed is some investigation on

> what the new intel macs use. Then one can edit the correct files so

> the build can verify that you are on a supported platform and thus

> compile the correct code.



Is there anything I (with my rather limited hackery) can do to determine this?



(It's no huge deal for me, the precompiled mac binaries seem to work

just fine... but if there's anything I can do to be helpful to the

overall cause...)



-r



&gt;

> Brian.

&gt;

>

&gt;




    




  

  
  
   
     
    

  

    intel mac build error
  


  

     user name

     2006-02-14 19:55:17
  


  

    

  


  

    
On 2/15/06, Ryan Raaum <ryan.raaumgmail.com&gt; wrote:



> Is there anything I (with my rather limited hackery) can do to determine this?



Well, to me the platform specific coro stuff is all, "Add 16 bytes to

the stack frame. Oh? That failed? Try 24 bytes...&quot; and so on. 



Steve and Olivier Ansaldi have recently been working on cleaning up

and isolating the coro code so now is a good time to talk to them

about it. Drop by the #io channel on freenode sometime and ask Steve

for some pointers.



> (It's no huge deal for me, the precompiled mac binaries seem to work

> just fine... but if there's anything I can do to be helpful to the

> overall cause...)



Jonathan.




    




  

  
  
   
     
    

  

    intel mac build error
  


  

     user name

     2006-02-14 20:13:12
  


  

    

  


  

    
> (It's no huge deal for me, the precompiled mac binaries seem to work

> just fine... but if there's anything I can do to be helpful to the

> overall cause...)



I tried a change based on other BSDs x86 implemention. The process gets

about two dozen or so frames into a locals_performOn_ before a call to the

lexer causes stack corruption. I tried the fix below early last week and

have had no time to investigate beyond. The problem happens&nbsp; as State it

initializing itself, perhaps when it loading from IoVMCode.c? Not sure

yet... but it is likely a that an adjustment for the proper size or range

of the stack as in the PPC implementation, will get it working.



If Steve can comment quickly on what the PPC implementation is doing, I

can do a fix/compile/test here then submit a patch.



#if defined(_BSD_PPC_SETJMP_H_)

&nbsp; &nbsp; // OSX/PPC

#define setjmp _setjmp

&nbsp;   // no need for signal masks

#define longjmp _longjmp

  ; &nbsp;  memset(((char *)stack) + stacksize - 64, 0, 64);   ; &nbsp; // very crude...

  ; &nbsp;  buf[0] = ((intptr_t)stack + stacksize - 64 + 15) & ~15;

&nbsp; &nbsp; &nbsp; buf[21] = (intptr_t)func;

#if defined(_BSD_I386_SETJMP_H)

&nbsp; &nbsp; &nbsp;  // Our Macintel.

 &nbsp; &nbsp;  buf[9] = (intptr_t)(stack + stacksize); // sets ESP

 ; &nbsp; &nbsp; buf[12] = (intptr_t)func; // sets EIP

...



On the same subject, anyone have experience with the mach_ and thread_

primitives in OS X? I would like to explore using the machine/user context

structures in conjunction with these functions and add simple

get/setcontext implementations. Last weekend, I discovered that

JavaScriptCore makes use of the ucontext_t and mcontext_t with their

garbage collector marks stack objects. (The

::markStackObjectsConservatively method of the Collector object.) The

trick is that thread_get_state() can not be called on the currently

executing thread. The implementation (an apple addition to the KJS code

base) contains i386 and ppc/ppc64 #defines and felt this might be nice

model for Io on OS X going forward. Any suggestions on how we might apply

this within the single-thread process of Io?





Dave



*



David Morford

Founder, Partner

Semanteme LLC

www.semanteme.com








    




  

  
  
   
     
    






 [1-5]











   
 





about | contact  Other archives ( Real Estate discussion Medical topics )