[lfs-dev] Testing for purity - Chapter 5
zarniwhoop at ntlworld.com
Sun Sep 2 18:48:44 PDT 2012
On Sat, Sep 01, 2012 at 05:01:07PM +0100, Jasmine Iwanek wrote:
> Results are from building ch5 glibc onwards in ch5 (not installed)
Sorry, I forgot I intended to reply to this.
> binutils build pass2: (superficial)
> cannot find -lstdc++
> potentially uses host c++ compiler
I don't know enough about the compiler to know if this is important
or not. Much of any gnu configure script is boilerplate, but I've
no idea how much is used (e.g. I frequently - when I look, which is
mostly in BLFS - see tests for fortran).
> Binary files
and others, which I've snipped.
In the past, did you ever do the similar exercise (using chapter 5
to build itself) ?
Anyway, if you are up for a variant of Greg's ICA, surely it is the
differences between chapter 6 built from chapter 5, and from itself,
which matter ?
In the past I used to use my farce scripts to compare chapter 6
(they're probably still in ~ken/ at lfs), but in the end I gave up
because (on x86_64) there were too many differences. To start with,
I had a list of files - some of which were in packages from BLFS
which I always build before booting - that always seemed to differ.
But somewhere, possibly around gcc-4.1, the list grew without any
good explanation. Using some program and option which I now forget
(probably an option to ld), I managed to disassemble the binaries -
it seemed that the code was *totally* different. I tentatively
marked this down as randomisation, i.e. various parts of the code
ended up at different places in the file, but without any way of
proving it. That was probably on x86_64, although I think Archaic
was the last person to run farce, probably on i686, and that he
saw similar 'inexplicable' differences.
das eine Mal als Tragödie, das andere Mal als Farce
More information about the lfs-dev