toolchain's output depends on toolchain used to build the compiler?

[Omer Zak]

I'd like to second Yedidyah's suggestion and re-express it as an
application of the "Lion in the desert" methodology.

The "desert division into halves by fences" can proceed along at least
two dimensions:

1. Reduction of the size of a module which compiles differently in the
two platforms (Yedidyah suggested to take out functions to their own
separate files).  First, by compiling a single function, then by taking
code out of it until you (Shachar) get the smallest code snippet that
builds differently in both systems.

2. Elimination of build steps as suspects - you (Shachar) already
eliminated the preprocessor.  You could eliminate the compiler proper by
building *.o files in one system, and building libraries + executables
in both systems & comparing them.  Similarly - for the linker and
binutils tools (of whose existence I was reminded by Oleg Goldshmidt).

                                                --- Omer


On Wed, 2010-02-10 at 08:37 +0200, Yedidyah Bar-David wrote:
> 3. Give us some more clues. This practically means a short example
> where this happens. Since you have the sources that compile differently,
> you might try to find out e.g. if taking just the function(s) that
> compiled differently and putting them in their own file still compiles
> them differently, and then post them (or something similar) (with your
> client's permission, if applicable). I know that gcc can optimize on
> a larger scope than a function, but this might work.
-- 
My own blog is at http://www.zak.co.il/tddpirate/

My opinions, as expressed in this E-mail message, are mine alone.
They do not represent the official policy of any organization with which
I may be affiliated in any way.
WARNING TO SPAMMERS:  at http://www.zak.co.il/spamwarning.html