Re: [Sbcl-devel] Core relocation (Was: WIN32 patch v.2)

[Juho S. <js...@va...>]

David Lichteblau <da...@li...> writes:
> Works now, at least on linux/ppc/cheneygc.
> Updated patch: http://www.lichteblau.com/blubba/relocate-3.diff

Great! Thanks.
 
> Questions:
> 
> It would be nice to check that relocation works in SBCL's test suite.
> Right now I am testing the patch manually using the TEST_START, _END
> environment variables, but obviously that should go away before
> committing the patch.  Is there a clever way to make SBCL start up with
> parts of its dynamic space blocked?

LD_PRELOAD a .so that does a suitable mmap in its constructor
function? 
 
> Also, what do you think about the "kludge"s in process_directory?

The new ordering/directory constraints seem quite acceptable. It's not
like we need to remain compatible with loads of existing software
which generates SBCL-format core files ;-)

But maybe the constraints should be explicitly checked in coreparse,
just in case.

-- 
Juho Snellman