stripping libgcj.so.4

Per Bothner per@bothner.com
Sun Sep 21 09:27:00 GMT 2003


Erik Poupaert wrote:

> I was looking at stripping libgcj.so itself, not really at stripping the executable.
> 
> Depending on how exactly the loading process goes, and to what extent it is cached
> (I'm not familiar with the precise details of this process), loading 8 MB instead
> of 40 MB of libgcj.so could conceivably be beneficial to startup time and memory
> footprint? Or is there a reason why it hardly matters?

I'm not too familiar with the details myself.  Clearly, you can't
completely strip the library of symbols, or applications won't be
able to dynamically link with it.  (An exception might be some shared
library schemes where the library is loaded at a fixed address.)

So the question is whether symbols that you *can* safely strip
(debugging symbols and local symbols) are in a separate
non-paged-in section than the symbols the linker needs.  I assume
they are in any well-design shared library scheme, but I don't
know for sure.
-- 
	--Per Bothner
per@bothner.com   http://per.bothner.com/




More information about the Java mailing list