mozart-1-3-0-fixes

[Boriss Mejias]

Good news.

Raph fixed the problem with the weak dictionaries fix (which was related to 
garbage collection) and committed to the trunk (r16950).

I merged r16910 and r16950 into mozart-1-3-0-fixes branch, so now weak 
dictionaries are correctly finalized and test suite is back again. It passes 
both on default and debug mode.

cheers
Boriss

Boriss Mejias wrote:
> After a deep study of the last commits done to the repository, I 
> realized that the test suite is broken due to the changes on Weak 
> Dictionaries. Those changes (rev16910) were meant for fixing Weak 
> Dictionaries termination, but they introduced some other bugs (see the 
> list of broken tests from the previous email).
> 
> The last patches on the marshaler are fine and they don't break anything.
> 
> So, I removed the changes related to Weak Dictionaries from 
> mozart-1-3-0-fixes, and added the fix for the marshaler, so the branch 
> is stable again.
> 
> Other changes merged into mozart-1-3-0-fixes are:
> 
> - The fix for using paths with spaces
> - Fix to IsOdd
> 
> Raph, can you have a look to your patch related to WeakDictionaries?
> 
> that's it for now.
> cheers
> Boriss
> 
> 
> Boriss Mejias wrote:
>> Kevin Glynn wrote:
>>> Boriss Mejias writes:
>>>  > Hi Kevin,
>>>  >  > I've been merging some of the changes I believe to be safe into 
>>> the  > mozart-1-3-0-fixes, so, there are some improvements already 
>>> there in  > comparison to 1.3.2. I haven't got time to run the test 
>>> suite with the fixes  > on the marshaling yet, so those are still 
>>> missing. Of course, I can merge your  > last changes with respect to 
>>> warnings on gcc 1.4.2 that you are working on  > right now.
>>>  >  > cheers, and thanks once again.
>>>  > Boriss
>>>  >
>>> No problem,
>>>
>>> Since I don't run Mozart or program much in C(++) I don't want to
>>> commit my stuff directly to subversion unless its a no-brainer.
>>>
>>> The warning stuff is a bit painful, and I think we will have to live
>>> with a lot of warnings at least until the next major release because
>>> they would require changes to signatures in mozart.h and mozart_cpi.hh
>>> and changing these could break people linking to pre-compiled code.
>>>
>>
>> mmm... as you said, it doesn't look like something we can just add in 
>> this minor release.
>>
>> I have also some concerns about the fix on the marshaler. The 
>> situation is the following.
>>
>> - Running on debug mode, Mozart passes the test-suite (with no dp), 
>> BUT, there is an emulator that crashes. This is probably a process child.
>>
>> - Without debug mode the test-suite is terrible. The following tests 
>> are known to crash: fd_fraction_all_entailed, fd_distribute_naive, 
>> fd_alpha_one, many other fd, schedule, fs, space_port and gump_scanner.
>> The test do not arrive to the point when it says FAILED, but it 
>> crashes in any of these tests, before proceeding.
>>
>> Any idea about why the test suite can pass when it is run on debug 
>> mode, and not it normal mode?
>>
>> cheers
>> Boriss
>>
>>> I'll send a patch amd more details soon, probably a day or so. Unless
>>> anyone has any insights and they want to discuss it sooner.
>>>
>>> k
>>>
>>>
>>>    
>>> _________________________________________________________________________________ 
>>>
>>> mozart-hackers mailing list                           
>>> mozart-hackers at mozart-oz.org      
>>> http://www.mozart-oz.org/mailman/listinfo/mozart-hackers
>>
>