Hi Jeff,

Sorry for the late reply.

On Aug 13, 2010, at 11:59 AM, Jeff Hemmelgarn wrote:

> Alright!
> 
> Thanks for all the help.  I was able to track down the problem and have some 
> changes for marshal.c that I need to get back to the repository.  There were 
> some assumptions about the sizes of types that were broken by going to 64-bit 
> .

Excellent!

> I don't know the procedures on this project.  Do I need to include tests for 
> the fix?  I am familiar with testunit but I haven't used rspec yet.  Where 
> and how would I include tests if I write them?  What documentation is normal 
> for the trouble ticket?  How do I get the changes into the repository?  
> Anything else I need to know?

Feel free to open a ticket on trac, describe the problem and attach your patch 
there. We will review and merge it once it's good.

We currently only use RubySpec as a regression test suite, so if you desire, 
looking at the pack specs and writing a new one if applicable might be a good 
idea too.

Laurent

On Aug 11, 2010, at 6:49 PM, Laurent Sansonetti wrote:
> 
>> Also, when debugging MacRuby it is better to use the version of libmacruby 
>> in the build directory, because the installer strips symbols during 
>> installation. From the build directory:
>> 
>> DYLD_LIBRARY_PATH=. gdb --args ./macruby ...
>> 
>> Also, the whole project is compiled with heavy optimizations which makes the 
>> debugging of certain things harder (inline functions, optimized variables, 
>> etc.). Sometimes it is better to debug a version of MacRuby with no 
>> optimization. You can build one using:
>> 
>> rake optz_level=0 macruby
>> 
>> HTH,
>> Laurent
>> 
>> On Aug 11, 2010, at 3:28 PM, Thibault Martin-Lagardette wrote:
>> 
>>> Hi,
>>> 
>>> Good to know you're motivated to get into hacking MacRuby :D
>>> 
>>> `rb_marshal_load` is an "API", supposed to be used by extensions for 
>>> example.
>>> If you look in `marshal.c`, you'll see that the "load" method is defined by 
>>> `marshal_load` (without the rb_ prefix):
>>>         rb_objc_define_method(*(VALUE *)rb_mMarshal, "load", marshal_load, 
>>> -1);
>>> 
>>> This means that when a ruby code calls Marshal.load, it will call this 
>>> method.
>>> To sum things up, I think you just need to set a breakpoint to 
>>> marshal_load, not rb_marshal_load :D
>>> 
>>> Let us know how things work out!
>>> -- 
>>> Thibault Martin-Lagardette
>>> 
>>> 
>>> 
>>> On Aug 11, 2010, at 13:33, Jeff Hemmelgarn wrote:
>>> 
>>>> Hello,
>>>> 
>>>> First I want to thank everyone involved in the project for their efforts 
>>>> in bringing MacRuby to us.  I love being able to use GCD with ruby!  I 
>>>> have been on vacation for a week and finally feel like touching a computer 
>>>> again so I thought I would look into doing some MacRuby development.
>>>> 
>>>> Since I have no experience with llvm and almost none with ObjectiveC, I 
>>>> decided to pick a low-priority, narrow focus bug to look into.  That 
>>>> should allow me to get used to the environment without going too far into 
>>>> the weeds. I pretty much randomly chose #502 Marshalling large integers 
>>>> returns wrong value.
>>>> 
>>>> I have downloaded and built both llvm and the latest code from the git 
>>>> mirror.
>>>> 
>>>> I have been trying to reproduce the problem with macirb under gdb.  I have 
>>>> some experience with gdb but I am getting confused trying to set a 
>>>> breakpoint in the Marshal package load function.  It looks like MacRuby is 
>>>> bypassing the globally visible function rb_marshal_load and using 
>>>> rb_objc_define_method to export a pointer to the file local function 
>>>> marshal_load.
>>>> 
>>>> The basics of the question is, how do I set a breakpoint in gdb for this 
>>>> function?  Is this problem a result of using llvm?  Will I end up needing 
>>>> to set a generic breakpoint in a dispatch routine with a condition looking 
>>>> for the selector for the load function?
>>>> 
>>>> Thank you in advance for any help!
>>>> 
>>>> Jeff Hemmelgarn
>>>> _______________________________________________
>>>> MacRuby-devel mailing list
>>>> [email protected]
>>>> http://lists.macosforge.org/mailman/listinfo.cgi/macruby-devel
>>> 
>>> _______________________________________________
>>> MacRuby-devel mailing list
>>> [email protected]
>>> http://lists.macosforge.org/mailman/listinfo.cgi/macruby-devel
>> 
>> _______________________________________________
>> MacRuby-devel mailing list
>> [email protected]
>> http://lists.macosforge.org/mailman/listinfo.cgi/macruby-devel
> 
> _______________________________________________
> MacRuby-devel mailing list
> [email protected]
> http://lists.macosforge.org/mailman/listinfo.cgi/macruby-devel

_______________________________________________
MacRuby-devel mailing list
[email protected]
http://lists.macosforge.org/mailman/listinfo.cgi/macruby-devel

Reply via email to