Linux JDK-1.5.0_12 (i386) unusable on amd64

[Alexander Leidinger]

Quoting Greg Lewis <glewis at eyesbeyond.com> (from Tue, 10 Jul 2007  
16:49:19 -0600):

> On Tue, Jul 10, 2007 at 09:13:34AM +0200, Alexander Leidinger wrote:
>> Quoting "Mikhail T." <mi+metlife at aldan.algebra.com> (from Mon, 09 Jul
>> 2007 15:04:01 -0400):
>>
>> >Hello!
>> >
>> >I just tried to build the regular jdk15 port on my amd64 machine. It
>> >relies on the Linux jdk15 to bootstrap.
>> >
>> >Unfortunately, all invocations of javac and java were crashing with a
>> >HotSpot stack (linux-sun-jdk-1.5.0.12,2).
>>
>>
>> >I'm unsure, where the blame lies -- with the Linuxulator, or the amd64
>> >(6.2-stable from June 7th) -- but something needs to be done...
>>
>> Can you please try with a more recent -stable? On Sunday I committed
>> some fixes which may or may not help (I don't remember if those fixes
>> helped on -current or not).
>>
>> I also have some more fixes (MFCs) in the queue, if you are interested
>> to help test them on -stable please tell me.
>
> I'll try it on 6.2-RELEASE-pmumble to make sure its working there.
>
> FWIW, I've had reports that linux-sun-jdk15 has been broken on
> -CURRENT/amd64 since around December, so MFC'ing changes may not
> actually improve the situation.  Can you try it on current to see
> if it works for you?

Regarding current it is important to know if this is with the default  
2.4 linux emulation, or with the 2.6 linux emulation.

Apart from that Roman knows about some thread "limitation" which hits  
us here (I let him comment upon this), and AFAIR the JDK also tries to  
use epoll which is not available yet... but I may misremember about  
the last part.

The big picture is: Linux programs start to depend more and more on  
stuff which is not emulated yet in any FreeBSD release. That's the  
reason why I try since ~1.5 years to get some people together to have  
a look at improving the linuxulator. Roman made some nice progress  
last year in the GSoC, and this year he will probably finish the last  
important missing parts (again as part of the GSoC). But I don't  
expect any release before 7.1 to contain all necessary pieces (but not  
enabled by default, I assume; this may have to wait until 8.0, but the  
decision has to be made when all the code is in the tree).

Bye,
Alexander.

-- 
When the only tool you have is a hammer, every problem starts to look
like a nail.

http://www.Leidinger.net    Alexander @ Leidinger.net: PGP ID = B0063FE7
http://www.FreeBSD.org       netchild @ FreeBSD.org  : PGP ID = 72077137