Re: [jamboost] Re: Use of objcopy prevents prelink to work on Linux

From: Rene Rivera (grafik.list_at_[hidden])
Date: 2005-07-14 11:50:10

Markus Schöpflin wrote:
> Rene Rivera wrote:
>>Markus Schöpflin wrote:
>>>Is there some way to disable the use of objcopy, besides 
>>>creating a new tools file?
> 
>>Hm, it doesn't look like it :-( It hadn't come up until now, it's easy 
>>enough to check for some option (--no-objcopy) on the invocation. Of 
>>course I wonder why is it that the link doesn't work for you given that 
>>all BBv1 does with objcopy is remove the 'readonly' flag from debug_str 
>>sections.
> 
> Nono, it's _pre_link not working. It seems that objcopy disturbs some order 
> prelink expects to find in the object file.
> 
> BTW, I need prelink to relocate my python extension module, to allow it to 
> cooperate with an external library that has a fixed location for a shared 
> memory segment, which unfortunately collides with the standard location 
> where the module is mapped when loaded by the python interpreter.

Ah, so that's why it hasn't come up. Not many people have a need for 
absolute object offsets. And when they do it's usually in C not C++. 
Well to help you out I've already added a "--no-objcopy" option check to 
gcc-tools.jam. SO at least you can get around it. We'll have to 
investigate if we really need objcopy in BBv2 in the future.

-- 
-- Grafik - Don't Assume Anything
-- Redshift Software, Inc. - http://redshift-software.com
-- rrivera/acm.org - grafik/redshift-software.com
-- 102708583/icq - grafikrobot/aim - Grafik/jabber.org