Re: Broken objc_allocateClassPair

[David Chisnall]

On 16 May 2013, at 05:48, Luboš Doležel <address@hidden> wrote:

> On 01/15/2013 04:13 PM, David Chisnall wrote:
>> On 14 Jan 2013, at 12:11, Luboš Doležel wrote:
>> 
>>> OK, great, I'll then ask the author of the patch to check if it works for 
>>> him :-)
>> 
>> It's in now, and there is a regression test that is registering 0xf0000 
>> selectors, which passes.
>> 
>> There was a bug in expanding sparse arrays, which meant that the empty 
>> children were always filled with empty root nodes.  This is now fixed, and 
>> means that we don't have to make any changes to the message lookup code for 
>> applications that use under 0xffff selectors.
>> 
>> David
> 
> Hi again,
> 
> any plans on making a release with all the fixes in trunk? :-)

Yes, I could probably branch an RC now actually.  I want to check that the 
optimisations build with the LLVM 3.3 release branch.  MIPS support will still 
be considered experimental in this release, but the message sending tests are 
passing for me on our MIPS system, except for the ones that involve exceptions 
(our linker doesn't correctly relocate the unwind sections yet...).

Please test the current trunk and report bugs - what is currently in trunk is 
likely to be the next release.

David

-- Sent from my STANTEC-ZEBRA