[Openmcl-devel] Trying to use AddressBook

[Gary Byers]

It's a bug.

The short version is that hash-tables with a (non-standard) :TEST
function of #'STRING= don't work. (I'm not sure about other non-standard
test functions.)

There are about a dozen places in "ccl:library;parse-ffi.lisp" that 
create such hash tables: about 4 places in the function PARSE-FFI,
and the remainder in PARSE-STANDARD-FFI-FILES.  Changing these to
use EQUAL (instead of STRING=) seems to work around the problems
with non-standard hash-table test functions.

(The first time that you run (PARSE-STANDARD-FFI-FILES :ADDRESSBOOK),
you'll get warnings to the effect that the .cdb files don't exist.
They don't: they're being created.  Running it a second time isn't
a bad idea; if the headers contained things like:

#define FOO 1
...
#define BAR (FOO+1)


there's a greater chance that #$BAR will work after a second pass,
because #$FOO will be defined.  I don't think that there's any
advantage to running PARSE-STANDARD-FFI-FILES more than twice.)


On Sat, 29 Oct 2005, Ralph Richard Cook wrote:

> I'm having some trouble. I (believe I) successfully made the AddressBook FFI; 
> I have a file about 4 megs big that looks like it has AddressBook constants, 
> includes etc. in it.
>
> From within OpenMCL I try
>
> CL-USER> (in-package :ccl)
> #<Package "CCL">
> CCL> (parse-standard-ffi-files :addressbook)
>
> and get
>
> value #<Unbound> is not of the expected type (OR STRING SYMBOL CHARACTER).
>   [Condition of type TYPE-ERROR]
>
> Restarts:
>  0: [ABORT] Abort handling SLIME request.
>  1: [ABORT-BREAK] Reset this process
>  2: [ABORT] Kill this process
>
> Backtrace:
>  0: (STRING 'NIL)
>  1: (CCL::STRING-START-END "ABPropertyType" #<Unbound> 0)
>  2: (STRING= "ABPropertyType" #<Unbound>)
>  3: (CCL::%HASH-PROBE #<HASH-TABLE :TEST STRING= size 0/60 #x8653426> 'NIL 
> #<HASH-TABLE-VECTOR  #x8653496>)
>  4: (GETHASH "ABPropertyType" 'NIL "ABPropertyType")
>  5: (CCL::FIND-OR-CREATE-FFI-TYPEDEF ':TYPE)
>  6: (CCL::PROCESS-FFI-TYPEDEF '(:TYPE ("/System/Library/Frameworks/ 
> AddressBook.framework/Headers/ABTypedefs.h" 34) "ABPropertyType" (:ENUM-REF 
> "_ABPropertyType")))
>  7: (CCL::PARSE-FFI #P"/usr/local/openmcl/ccl/darwin-headers/ 
> addressbook/C/System/Library/Frameworks/AddressBook.framework/Headers/ 
> AddressBook.ffi")
>  8: (CCL::PARSE-STANDARD-FFI-FILES ':ADDRESSBOOK)
>  9: (CCL::CALL-CHECK-REGS 'CCL::PARSE-STANDARD-FFI-FILES)
> --more--
>
> After looking at parse-ffi.lisp I try
>
> CCL> (describe '*ffi-typedefs*)
> Symbol: *FFI-TYPEDEFS*
> Special Variable
> INTERNAL in package: #<Package "CCL">
> Print name: "*FFI-TYPEDEFS*"
> Value: #<Unbound>
> Function: #<Unbound>
> Plist: NIL
> ; No value
>
>
> Any ideas?
>
> Thanks
>
> On Oct 23, 2005, at 8:47 PM, Gary Byers wrote:
>
>> AddressBook is likely something like WebKit (a set of Cocoa classes
>> that are large enough and domain-specific enough to be distributed as
>> their own framework.)
>> 
>> The good news is that there's a (very simple) example of using WebKit
>> in the OpenMCL 1.0 distribution, and the steps involved in using 
>> AddressBook shold be similar.  The bad news is that there's some
>> one-time setup involved and a little bit of startup work to be
>> done in any code that uses AddressBook (or WebKit, or ...) under
>> the Cocoa Bridge.
>> 
>> The one-time setup involves generating interfaces (.cdb files)
>> describing the classes and methods in the AddressBook headers,
>> so that it'd be possible to say something like:
>> 
> ....
>> 
>> If this makes sense so far, great, and if you want to pursue this
>> on your own please let the list know if you're successful or if
>> you encounter further problems.
>> 
>> 
>
>