[m-rev.] for review: improve support for using frameworks on Mac OS X

[Julien Fischer]

On Wed, 6 Oct 2010, Ian MacLarty wrote:

> On Wed, Oct 06, 2010 at 01:21:14PM +1100, Julien Fischer wrote:
>>
>> For review by Ian.
>>
>> Improve support for building and linking against frameworks on Mac OS X
>> by providing mmc command line options that allow which framework(s) to
>> build against to be specified directly.
>>
>> compiler/options.m:
>> 	Add a new option, --framework, that causes an executable
>> 	or library to be linked against the framework specified
>> 	by the argument.
>>
>> 	Add new option, --framework-directory, that appends a new
>> 	framework search directory to the current set of framework
>> 	search directories.  (The short for of the option is -F;
>
> s/for/form/

Fixed.

>
>> 	the reason for using this is that if the linker emits warnings
>> 	about the framework search directories, then it will refer
>> 	to the option by that name.)
>>
>> compiler/compile_target_code.m:
>> 	Pass the framework search directories to both the C compiler
>> 	and the linker.  (They need to go the C compiler so that it
>
> to go to

Fixed.

>> Index: compiler/compile_target_code.m
>> ===================================================================
>> RCS file: /home/mercury/mercury1/repository/mercury/compiler/compile_target_code.m,v
>> retrieving revision 1.160
>> diff -u -r1.160 compile_target_code.m
>> --- compiler/compile_target_code.m	30 Sep 2010 07:23:31 -0000	1.160
>> +++ compiler/compile_target_code.m	6 Oct 2010 02:13:46 -0000
>> @@ -437,6 +437,9 @@
>>          C_Incl_Dirs),
>>      InclOpt = string.append_list(list.condense(list.map(
>>          (func(C_INCL) = ["-I", quote_arg(C_INCL), " "]), C_Incl_Dirs))),
>> +
>> +    get_framework_directories(Globals, FrameworkInclOpt),
>> +
>>      globals.lookup_bool_option(Globals, highlevel_code, HighLevelCode),
>>      (
>>          HighLevelCode = yes,
>> @@ -889,7 +892,8 @@
>>      % listed after CFLAGS.
>>      %
>>      string.append_list([
>> -        SubDirInclOpt, InclOpt,
>> +        SubDirInclOpt, InclOpt, " ",
>> +        FrameworkInclOpt, " ",
>>          OptimizeOpt, " ",
>>          HighLevelCodeOpt,
>>          NestedFunctionsOpt, @@ -1885,6 +1889,9 @@
>>      get_runtime_library_path_opts(Globals, LinkTargetType,
>>          RpathFlagOpt, RpathSepOpt, RpathOpts),
>>
>> +    % Set up any framework search paths.
>> +    get_framework_directories(Globals, FrameworkDirectories),
>> +
>>      % Set up the install name for shared libraries.
>>      globals.lookup_bool_option(Globals, shlib_linker_use_install_name,
>>          UseInstallName),
>> @@ -1913,6 +1920,7 @@
>>          globals.lookup_string_option(Globals, TraceFlagsOpt, TraceOpts)
>>      ),
>>
>> +    get_frameworks(Globals, Frameworks),
>>      get_link_libraries(Globals, MaybeLinkLibraries, !IO),
>>      globals.lookup_string_option(Globals, linker_opt_separator, LinkOptSep),
>>      (
>> @@ -1961,8 +1969,10 @@
>>                      LinkOptSep, " ",
>>                      LinkLibraryDirectories, " ",
>>                      RpathOpts, " ",
>> +                    FrameworkDirectories, " ",
>>                      InstallNameOpt, " ",
>>                      DebugOpts, " ",
>> +                    Frameworks, " ",
>>                      LDFlags, " ",
>>                      LinkLibraries, " ",
>>                      MercuryStdLibs, " ",
>> @@ -2400,6 +2410,34 @@
>>          MadeSymlinkOrCopy = no
>>      ).
>>
>> +:- pred get_framework_directories(globals::in, string::out) is det.
>> +
>> +get_framework_directories(Globals, FrameworkDirs) :-
>> +    globals.lookup_accumulating_option(Globals, framework_directories,
>> +        FrameworkDirs0),
>> +    (
>> +        % Short-circuit the common case.
>> +        FrameworkDirs0 = [],
>> +        FrameworkDirs = ""
>
> Is this optimisation really necessary?

No - when I started working on this change there was a bit more going on
in the other branch.  I've removed it the above (in both cases).

> I mean this code is only going
> to get called once per module isn't it?

Yes.

Julien.
--------------------------------------------------------------------------
mercury-reviews mailing list
Post messages to:       mercury-reviews at csse.unimelb.edu.au
Administrative Queries: owner-mercury-reviews at csse.unimelb.edu.au
Subscriptions:          mercury-reviews-request at csse.unimelb.edu.au
--------------------------------------------------------------------------