icedtea6 1.11 branching on 2012-01-06

[Omair Majid]

On 01/06/2012 05:10 AM, Andrew Haley wrote:
> On 01/06/2012 07:16 AM, Dr Andrew John Hughes wrote:
>> On 15:48 Thu 05 Jan     , Matthias Klose wrote:
>>> On 12/23/2011 04:31 PM, Omair Majid wrote:
>>>> Hi all,
>>>>
>>>> As noted elsewhere, a number of important features and fixes have been added to
>>>> icedtea6 that are not in any release. This seems like a good time as any to try
>>>> and get a release out. To that end, I am planning to branch icedtea6 1.11 on
>>>> 2012-01-06. I dont have an exact release date in mind right now, but it probably
>>>> will be at least a week after that.
>>>
>>> there's a regression: shark is completely broken. seen on the buildbot
>>> http://builder.classpath.org/icedtea/buildbot/builders/icedtea6-squeeze-x86_64-quick-shark/builds/188
>>> using llvm-2.7, and checked with llvm-2.9 and llvm-3.0 as well. If not fixed,
>>> please mention it in the release notes.
>>
>> Do you plan to fix it?
>
> As I understand it, LLVM-JIT is horribly broken while they rewrite
> the JIT back-end.  If that's true, we don't want to update Shark
> until they're done.  However, I don't read the LLVM lists, so this
> may all be nonsense.

Does that mean shark works fine with older versions of llvm? Would that 
be llvm-2.6? I ask because I have vague memories of testing shark with 
llvm-2.6 which contained enough bugs that made shark not work. And 
unless I am mistake things were working better with llvm-2.7. Have there 
been changes to shark that have made it not work on 2.7?

I am trying to figure out what level of shark support should claim to 
offer in icedtea6 1.11 ("works if the stars are aligned just right" vs 
"nope, broken").

Thanks,
Omair