August 2009 - MacRuby-devel - lists.macosforge.org

experimental branch: status update
by Laurent Sansonetti 06 Aug '09

06 Aug '09

Headlines: - macirb should work as before. The local variable bug has been fixed. - macrake should work as before. Running HotCocoa projects should work, you can even build MacRuby with macrake. - new YAML module, API compatible with syck, was added. It is still under development, but it's currently good enough for most uses, including HotCocoa rakefile tasks. This is the last experimental branch status update, we accomplished I think all the goals required to merge the branch into trunk. There will be more status updates after, but they will focus on trunk and the 0.5 release objectives :) I will proceed with the merge tomorrow at 3PM California time (midnight Amsterdam time, 7AM Tokyo time). Feel free to commit before, but please hold off your commits at that time :) Changes: - fixed an incompatibility that showed up in switching between libedit and GNU readline - added support to delete environment variables using ENV[]=nil - fixed the clean task to remove binary build producfs too. - fixed GC problems in oniguruma: make sure st.c tables are retained/ released appropriately since they use GC memory - fixed a problem in IO#gets where the stream wasn't marked as EOF after reading the last separator-terminated line - implemented Thread.start/fork - added support for creation of Binding local variables - fixed a bug in the way we compile return-from-block handlers - fixed a bug in IO#reopen - #fork is now raising a "not yet supported" exception, because it doesn't work well with CF and libauto. We will try to support it, but later. - fixed String#inspect to escape some characters - ported Rational and Complex to the new runtime APIs - various unicode/bytestring fixes - fixed AOT compilation of keep locals. - fixed bytestring -> path conversion - keep IO streams that should never be closed into a static array to avoid them being collected - fixed a bug in the fast aref primitive: convert fixnum argument to long and not int - fix a bug in the dispatcher where calling a method with an empty splat array wasn't dispatching a zero arity method - optimize numeric coerce dispatch calls - overwrite -[NSObject description] in every new subclass that calls #to_s + optimized other overloaded methods - changed the way %s is implemented to behave like the ruby spec and send #to_s - updated the stringscanner specs to be 1.9.2 compatible - updated the stringio specs to be 1.9.2 compatible - RubySpec was updated from upstream - implemented conditional assignment of class variables - work around a crash while raising an objc exception from a ruby one - ported the compiler to 32-bit (including the floating point optimization) - removed the negative-index feature of Readline::HISTORY (of libedit) to behave like readline - optimized Math.sqrt - fixed super within a method that has a splat argument - added Integer#ord - a YAML module has been written, based on the libyaml C library. It is still under development but load and dump should work. See the previous e-mail on the mailing-list for more information. - a pure Ruby stringio module is under development Laurent

7 16

How to use custom C struct's?
by Clay Bridges 06 Aug '09

06 Aug '09

Hi all, I'm writing an iPhone app. I made the audacious step of trying to test my ObjC model classes using MacRuby. I create and load a dynlib ("bundle") of these classes. So far, so good. However, I have a C struct that I use all over the place: struct MapPoint { int row; int col; }; typedef struct MapPoint MapPoint; Obviously, when I have a method that returns that type, MacRuby doesn't like it. That is, # ruby my_objc_object.mapPoint gives this error: RuntimeError: unrecognized octype `{MapPoint=ii}' The google didn't yield much guidance on this. Before I started a deep-dive on the MacRuby source, and/or the standard ruby way to handle this sort of thing, I thought I would ask a couple of questions: 1) Any easy advice? 2) Is this different in MacRuby than in ruby proper? 3) Any pointers into the MacRuby source that might help me? Thanks, Clay Bridges

5 8

Information about the new YAML module
by Patrick Thomson 06 Aug '09

06 Aug '09

Hi everyone, I've just pushed my work on a new YAML module to the experimental branch (revision 2184). Rather than being backed by the old syck code that 1.8/1.9 use, this is backed by the libYAML library (BSD licensed). As of right now it appears to be about 4x slower than 1.9's YAML module, but there's a lot of room for optimizing my code. This new YAML module aims to be wholly compatible with the old API inside yaml.rb. *Please *test this module out, and let me know if you find any inconsistencies or false results. As of right now it passes the RubySpecs for YAML::dump() and YAML::load(). (Oh, and sorry for spamming the macruby-changes list; git-svn applied something like 40 commits at one time. Oops!) -- Patrick Thomson

5 8

MacRuby Build Issues
by John-Paul Bader 05 Aug '09

05 Aug '09

-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 Hey guys, i'm relatively new in the world of MacRuby but I'm hooked already. I tried to build it from source (trunk and experimental) but didn't have any luck with it. The experimental branch exits here: … "llvm::Intrinsic::getDeclaration(llvm::Module*, llvm::Intrinsic::ID, llvm::Type const**, unsigned int)", referenced from: RoxorCompiler::compile_bs_struct_new(rb_vm_bs_boxed*) in bridgesupport.o RoxorCompiler::compile_landing_pad_header(std::type_info const&)in compiler.o RoxorCompiler::compile_landing_pad_header(std::type_info const&)in compiler.o RoxorCompiler::compile_landing_pad_header(std::type_info const&)in compiler.o "llvm::LoadInst::LoadInst(llvm::Value*, char const*, bool, llvm::Instruction*)", referenced from: RoxorAOTCompiler::compile_sel(objc_selector*, bool) in compiler.o RoxorAOTCompiler::gen_slot_cache(unsigned long)in compiler.o ld: symbol(s) not found for architecture x86_64 collect2: ld returned 1 exit status lipo: can't open input file: /var/folders/YT/YT9QPpMNF4KUf1I+NyCXyk++ +TI/-Tmp-//ccG7VEk8.out (No such file or directory) rake aborted! Command failed with status (1): [/usr/bin/g++ -I. -I./include -I./onig - -I/u...] I installed the recommended llvm version and I tried to rake clean before every attempt The trunk exits here: … gcc -I. -I../../.ext/include/universal-darwin9.0 -I..//.././include - I..//.././ext/readline -DRUBY_EXTCONF_H=\"extconf.h\" -fno-common - fno-common -pipe -O2 -g -Wall -Wno-parentheses -arch i386 -arch x86_64 - -o readline.o -c readline.c gcc -dynamic -bundle -undefined suppress -flat_namespace -arch i386 - arch x86_64 -o ../../.ext/universal-darwin9.0/readline.bundle readline.o -L. -L../.. -arch i386 -arch x86_64 -lmacruby -lreadline - -lncurses compiling ripper ../../miniruby -I'../..' -I'..//.././lib' -I'../../.ext/universal- darwin9.0' -I'../../.ext/common' -I./- -I'..//.././ext' - rpurelib.rb ..//.././ext/ripper/tools/preproc.rb ..//.././parse.y -- output=ripper.y bison -t -v -oy.tab.c ripper.y ripper.y:2920.25-26: $$ für Regelmitte bei $6 von „primary“ hat keinen deklarierten Typ make: *** [ripper.c] Error 1 rake aborted! Command failed with status (1): [GC_DISABLE=1 ./miniruby -I./lib - I.ext/com...] /usr/local/src/MacRuby-trunk/Rakefile:591 (See full trace by running task with --trace) My shell has a german locale. Forgive me ;) I don't really have a clue what I could do to fix it. I'm following the list for a few days now and didn't here any reports of build issues. Any ideas? Kind regards, John -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.9 (Darwin) iEYEARECAAYFAkp5Q4YACgkQK6xKqn9SkYOCIwCgvi1ZjBjPyhi7b/+VywdBvhFQ nOcAnR1IvUhw/QfLQlxd18SlT5eE2syR =MkN9 -----END PGP SIGNATURE-----

2 2

NSNumber boolean spec
by dan sinclair 04 Aug '09

04 Aug '09

The attached patch gets the NSNumber boolean conversion spec working. I'm not sure if the change is correct but I changed the spec from should != to should_not == and it appears to be working correctly now. dan

2 1

Add Integer#ord
by dan sinclair 04 Aug '09

04 Aug '09

Attached patch ports Integer#ord from Ruby 1.9. dan

2 1

Miniruby Causing Trouble with SVN Edge
by Maxwell Hawkins 04 Aug '09

04 Aug '09

I've been trying to compile 0.5 from SVN with no luck. I installed LLVM revision 72741 as outlined in the readme but when I ran MacRuby's rake file I got the following error: > ../../miniruby -I../.. -I../../lib -r rbconfig -e > "RbConfig::CONFIG['libdir'] = '../..'; require './extconf.rb'" > Assertion failed: (castIsValid(getOpcode(), S, Ty) && "Illegal > BitCast"), function BitCastInst, file Instructions.cpp, line 2493. Thinking my environment might be the problem, I installed a fresh copy of Leopard and reinstalled LLVM. I still get the same error message. Further investigation reveals that the problem occurs when rbconfig is required by miniruby in line 325 of builder.rake. It throws the error when when it encounters the "==" in "RUBY_VERSION == "1.9.0"" at the top of rbconfig.rb. I encounter similar problems (and the same error message) concatenating strings (-e "'string' + 'string'") with miniruby. Both experiments share a similar gdb backtrace: > #0 0x9309de42 in __kill () > #1 0x9309de34 in kill$UNIX2003 () > #2 0x9311023a in raise () > #3 0x9311c679 in abort () > #4 0x931113db in __assert_rtn () > #5 0x00969cff in llvm::BitCastInst::BitCastInst () at vm.h:582 > #6 0x0012efa9 in RoxorCompiler::compile_double_coercion > (this=0xa029200, val=0xe6578c, mask=0xe65fe8, fallback_bb=0xe65ee0, > f=0xe641f0) at compiler.cpp:1929 > #7 0x00138231 in RoxorCompiler::compile_optimized_dispatch_call > (this=0xa029200, sel=<value temporarily unavailable, due to > optimizations>, argc=1, params=@0xbfffcb4c) at compiler.cpp:2254 > #8 0x00149276 in RoxorCompiler::compile_node (this=0xa029200, > node=0x181a150) at compiler.cpp:3813 > #9 0x0013f5e6 in RoxorCompiler::compile_node (this=0xa029200, > node=0x181a070) at compiler.cpp:4029 > #10 0x001427f0 in RoxorCompiler::compile_node (this=0xa029200, > node=0x1811fe0) at compiler.cpp:3598 > #11 0x00146833 in RoxorCompiler::compile_node (this=0xa029200, > node=0x1859c00) at compiler.cpp:2876 > #12 0x0014f770 in RoxorCompiler::compile_main_function > (this=0xa029200, node=0x1859c00) at compiler.cpp:4717 > #13 0x0017ca65 in rb_vm_run (fname=0x0, node=0x0, binding=0x0, > inside_eval=false) at vm.cpp:4132 > #14 0x00040a01 in rb_load (fname=<value temporarily unavailable, due > to optimizations>, wrap=0) at load.c:293 > #15 0x000415c3 in rb_require_safe (fname=25483856, safe=0) at load.c: > 488 > #16 0x001701a8 in rb_vm_dispatch (cache=0xe61b30, self=25258368, > sel=0x3cbf, block=0x0, opt=0 '\0', argc=<value temporarily > unavailable, due to optimizations>) at vm.cpp:2026 > #17 0x0a880062 in ?? () > #18 0x0003879b in ruby_run_node (n=0x1816980) at eval.c:198 > #19 0x0011d2a8 in main (argc=3, argv=0xe04c30, envp=0xbffff724) at > main.cpp:38 Is there any way I can get this edge release working? I'm on a MacBook Pro with a 2.16GHz Core Duo running 10.5.7. Thank you, Max Hawkins

3 3

RubySpec updated
by Eloy Duran 03 Aug '09

03 Aug '09

Hi, I'm un-jet lagging a bit, so I thought I'd update the ruby specs again. We are now passing: 18160 examples. @Matt: Great work on the StringScanner! Could you please make sure the specs run on 1.8 as well? Currently 4 fail: $ mspec -B ruby.1.8.mspec library/stringscanner StringScanner#getch is multi-byte character sensitive FAILED StringScanner#getch should keep the encoding ERROR StringScanner#initialize_copy is a private method FAILED StringScanner#initialize is a private method FAILED @Laurent: There were quite some changes to complex that you might want to look into since you recently worked on those :) Cheers, Eloy

5 10

LLVM Build Convenience
by Josh Ballanco 03 Aug '09

03 Aug '09

Hey all, I got tired of constantly having to rebuild/reinstall LLVM, so I added a facility to scrape the commands from the README and offer to run them if llvm-config isn't found. I've attached the diff if anyone else is interested (don't know that I'd recommend this for general use, though). It's a git diff, but "git apply llvm_build.diff" will work even if you've checked out the repo using svn. Cheers, Josh

3 3