The secret sauce in Tango's AAs (Re: Garbage collector...)

[Bill Baxter]

Christian Kamm wrote:
> Bill Baxter wrote:
> 
>> Well a good start would be to fix phobos' AA implementation so that it
>> doesn't generate so many false positives.  Tango improves on it some,
>> but I wonder if Tango does ok on the test case here:
>> http://d.puremagic.com/issues/show_bug.cgi?id=1561
> 
> I just modified the code you provided for tango and ran it: Tango's AAs seem
> not to exhibit this problem at all. (check bug for details)
> 
> Christian


Thanks for trying that out.  I don't understand why it would work on 
Tango, though. There must be something different happening from what I 
guessed originally, because keysize and valuesize should both be equal 
to (void*).sizeof on a 32 bit machine, and thus Tango shouldn't be 
setting the NO_SCAN flag.

Actually looking closer, it appears that the hash value used by DMD 
doesn't even look pointer-like.   Actually it's just an identity function:
     int x = 99;
     typeid(int).getHash(cast(void*)&x) // == 99

In the example we're just hashing values 0-999, so those shouldn't be 
mistaken as pointers.

Can anyone explain why Tango's AA's work right when Phobos's don't here?


--bb