the original script came from binutils and we simply adapted it.

Yes, but that script is meant for userspace things and isn't
the right thing for your environment.

I tried to simplify but I was unable to predict all the gcc created sections for various gcc flags (esp -O2), so I just put all that stuff back.

GCC won't create too many sections, just a few, unless you
do something terribly wrong.

I'll try another pass at another time, when I can test all scenarios.

Good to hear.  I consider this patch reversion to be a
regression though.

There is a huge degree of risk playing here, you can imagine how hard an issues from a dropped section would be hard to detect.

I tend to fix such issues, they're not hard to detect at all
(almost all of the time) -- dropped sections lead to hard
crashes really easily ;-)

Figuring out what is wrong is another thing -- often it is
*not* the linker script but something else btw, you just
get lucky with a certain script and a certain binutils
version but unlucky with another combo.

Perhaps ld has options to warn when badness occurs.

Yeah it can do many of those things.


Again -- do you have a testcase for the failure that made
you revert this patch?


Segher


_______________________________________________
Xen-ppc-devel mailing list
Xen-ppc-devel@lists.xensource.com
http://lists.xensource.com/xen-ppc-devel

Reply via email to