2.6.16 for linuxulator & 7.0 release

[Alexander Leidinger]

Quoting Divacky Roman <xdivac02 at stud.fit.vutbr.cz> (from Mon, 19 Mar  
2007 19:35:44 +0100):

> On Mon, Mar 19, 2007 at 06:06:18PM +0100, Alexander Leidinger wrote:
>> Quoting Divacky Roman <xdivac02 at stud.fit.vutbr.cz> (from Mon, 19 Mar
>> 2007 17:49:07 +0100):
>>
>> >On Fri, Mar 16, 2007 at 12:48:12PM -0500, Dan Nelson wrote:
>> >>In the last episode (Mar 16), Alexander Leidinger said:
>> >>> Quoting Divacky Roman <xdivac02 at stud.fit.vutbr.cz> (from Fri, 16 Mar
>> >>> 2007 13:26:58 +0100):
>> >>>
>> >>> >as java is regarded I think its not a big problem to patch java
>> >>> >shell script to set emulation to 2.4 etc. but I agree that 2.6 is
>> >>> >not ready for 7.0R
>> >>>
>> >>> This would require to run every java programs as root. Very impractical.
>> >>
>> >>Wouldn't setting LD_ASSUME_KERNEL=2.4.20 (or any 2.4.x release)
>> >>suffice?
>> >
>> >no... we act differenly in the kernel depending on how
>> >compat.linux.osrelease
>> >is set... setting LD_ASSUME_KERNEL wont change a thing and it WILL
>> >SCREW THINGS
>> >
>> >dont do this
>>
>> Even when we set it to .20 as he suggested instead to .2 as I
>> suggested? With .20 we get the NPTL stuff and maybe not the epoll
>> stuff in the jdk.
>
> I was talking about setting LD_ASSUME_KERNEL.. this DOES NOT work in  
>  our current
> linuxulator and will BREAK things.
>
> the linux_use26() function works on the sysctl   
> compat.linux.osrelease and doesnt care
> about LD_ASSUME_KERNEL so kernel things its running 2.6 and userland  
>  thinks its running
> 2.4 -> things must break.

It depends on what is used. 2.6 in the kernel changes how some stuff  
is handled. If LD_...=2.4.20 expect that the kernel behaves in those  
areas like 2.6.16 and at the same times disallows the use of epoll()  
and other stuff we don't want to be used, this would be an option. But  
this depends upon how java determines how to use epoll() and upon the  
expectations of glibc with LD_...=2.4.20. I didn't verified this but  
it is easy to verify.

A thought which just crossed my mind is: is an user able to panic the  
kernel when 2.6.16 is in effect in the kernel and LD_... is set to  
2.4.x and vice versa?

Bye,
Alexander.

-- 
  La Brea Tar Pits
  As seen on the tar channel

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