bug in scm_arrray_handle_[srfi tag]_elements

bug in scm_arrray_handle_[srfi tag]_elements

[Daniel Llorens del Río]

Hello,

That function and the _writable_elements() version are broken for  
shared arrays because they return a pointer to the beginning of the  
enclosing array instead of a pointer to their own beginning.

This patch against master fixes the issue for me. Cf. the  
implementation of scm_array_handle_uniform_writable_elements() in  
uniform.c.

---
  libguile/srfi-4.c |    4 ++--
  1 files changed, 2 insertions(+), 2 deletions(-)

diff --git a/libguile/srfi-4.c b/libguile/srfi-4.c
index 4b7eadc..f38fbcf 100644
--- a/libguile/srfi-4.c
+++ b/libguile/srfi-4.c
@@ -119,13 +119,13 @@
     
{                                                                     \
      if (h->element_type != ETYPE  
(TAG))                                 \
        scm_wrong_type_arg_msg (NULL, 0, h->array, #tag  
"vector");        \
-    return h->elements;                               \
+    return h->elements+h->base*scm_array_handle_uniform_element_size 
(h);\
    }                                                                    
   \
    ctype* scm_array_handle_##tag##_writable_elements  
(scm_t_array_handle *h) \
     
{                                                                     \
      if (h->element_type != ETYPE  
(TAG))                                 \
        scm_wrong_type_arg_msg (NULL, 0, h->array, #tag  
"vector");        \
-    return h->writable_elements;                      \
+    return h->writable_elements+h- 
 >base*scm_array_handle_uniform_element_size(h); \
    }                                                                    
   \
    const ctype *scm_##tag##vector_elements (SCM  
uvec,                    \
                                             scm_t_array_handle  
*h,       \
-- 

Thanks,

	Daniel

Re: bug in scm_arrray_handle_[srfi tag]_elements

[Andy Wingo]

Hi Daniel,

On Mon 11 Jan 2010 19:49, Daniel Llorens del Río <daniel.llorens@bluewin.ch> writes:

> That function and the _writable_elements() version are broken for shared
> arrays because they return a pointer to the beginning of the  enclosing
> array instead of a pointer to their own beginning.

I've fixed the issues compiling e.g. #@2(1 2 3) and the like, and
applied this patch as well. Thanks for the reports.

Anything else? :)

Andy
-- 
http://wingolog.org/

Re: bug in scm_arrray_handle_[srfi tag]_elements

[Daniel Llorens del Río]

On 12 Jan, 2010, at 0:22, Andy Wingo wrote:

> I've fixed the issues compiling e.g. #@2(1 2 3) and the like,

Thanks for the great support.

> and applied this patch as well. Thanks for the reports.
>
> Anything else? :)

Yeah… your patch is wrong :)

-    return h- 
 >elements;                                                 \
+    return ((const ctype*) h->elements) + h- 
 >base;                      \

I wrote it this way at first, too, but it doesn't work for c32/c64  
because ctype for those is float/double. Calling the element size  
function as I did is clunky, since the type is already known, but  
maybe one could access the table in uniform.c directly or something.

Regards,

	Daniel

Re: bug in scm_arrray_handle_[srfi tag]_elements

[Daniel Llorens del Río]

On 12 Jan, 2010, at 0:49, Daniel Llorens del Río wrote:

> -    return h- 
> >elements;                                                 \
> +    return ((const ctype*) h->elements) + h- 
> >base;                      \
>
> I wrote it this way at first, too, but it doesn't work for c32/c64  
> because ctype for those is float/double.

Or maybe hardcode the sizes in the macro call, it's not like they are  
not already in two places.

Regards,

	Daniel.

Re: bug in scm_arrray_handle_[srfi tag]_elements

[Andy Wingo]

On Tue 12 Jan 2010 00:49, Daniel Llorens del Río <daniel.llorens@bluewin.ch> writes:

> On 12 Jan, 2010, at 0:22, Andy Wingo wrote:
>
>> Anything else? :)
>
> Yeah… your patch is wrong :)

Damn. I actually was thinking of this as I woke up, before I got to your
message.

I gave it another go; please take a look at it :)

Cheers,

Andy
-- 
http://wingolog.org/

Re: bug in scm_arrray_handle_[srfi tag]_elements

[Daniel Llorens del Río]

On 12 Jan, 2010, at 20:16, Andy Wingo wrote:

> I gave it another go; please take a look at it :)

Looks good! I had not noticed the other two cases.

Thanks,

	Daniel