steve@watchfield.com writes:
> Date: Mon, 25 Oct 2004 00:43:57 +0000 (UTC)
>
> Hi Aaron and Philippe,
>
> I am very interested in testing all aspects of Poplog on Darwin
Do you have Darwin running on a PC?
If not you can't do any testing till it is ported to PPC.
> including Common Lisp, Standard ML and Prolog.
> However, there are no instructions for testing or installation or
> anything - and this is off-putting.
I guess I should have checked before installing it here
http://www.cs.bham.ac.uk/research/poplog/darwin+pc/
I just assumed there were instructions in the tar file.
> Now I may be the only one
> who thinks like this - but I reckon we need a few paragraphs explaining
> to would-be testers what is required.
>
> [1] How to install AND uninstall.
> [2] How to configure and get started.
> [3] What kind of things need testing.
> [4] How and who to report problems to. Ideally an HTML form to
> fill in or, less work, a report template.
>
> After all, no point in doing anything unless [4] is crystal clear!
Agreed.
My guess is that it just copies a linux-poplog tree. So if you untar
it into something like this (or whatever the Darwin equivalent is):
/usr/local/darwin-poplog
then you set the environment variable usepop to be a
/usr/local/darwin-poplog/current.poplog
Then set all other variables by sourcing
$usepop/pop/com/poplog.sh
or if you prefer a csh/tcsh shell
$usepop/pop/com/poplog
I am assuming that under Darwin you can get a command line that
works as in linux. If not all the startup shell scripts will have to
be replaced by something else.
> It may also be worth adding that the release notes,
> http://www.cs.bham.ac.uk/research/poplog/darwin+pc/PoplogDarwinReleaseNotes.txt,
> are difficult to follow in places. In particular:
>
> >This Port was not completed because some generated assembler
> >instructions for Darwin were not accepted. However, since the
> >main goal was reached (to modify the Linux-Poplog
> >system calls so they were recognized by the Darwin Kernel),
> >it was decided to ignore this problem and to continue with the next
> >stages to achieve the final (and more important) objective (to
> >port poplog into Mac OS X).
I seem to recall this was explained in one of the other files or an
earlier message where a temporary work-around was adopted, namely
replacing a shift with an arithmetic operation, or something like
that.
But as I know nothing about Darwin (had have no assmbler programming
experience!) I can't add anything.
> It is unclear to me how one can continue with the port with this
> assembler problem in place. Does this mean that the offending assembler
> will be rewritten by hand?
I took it that the system-building sources had been altered to
generate the less efficient version.
I noticed that the tar file contained a lot of newly built saved
images. They would have been difficult to produce if the system were
not essentially working OK. That leaves the problem of testing the X
interface, external load, etc.
An important question is:
Does this problem arise ONLY when assembler files are being
generated in order to rebuild poplog, or is it also a problem
in the incremental compiler?
I assume not, since the code for that should be the same whether
it's Darwin or Linux that's running, apart from a few system calls.
In that case I wonder whether the problem is actually a bug in
the Darwin PC assembler.
Perhaps that can be checked by seeing whether the corresponding
C code also generates a bug when compiled under Darwin on PC.
Poplog has revealed bugs in many assemblers and linkers in the past.
If Darwin on PC is not used much (I have still not heard of anyone
using it) that may mean that it has lots of undetected bugs!
Aaron
====
Aaron Sloman, ( http://www.cs.bham.ac.uk/~axs/ )
PAPERS: http://www.cs.bham.ac.uk/research/cogaff/ (also talks in /talks )
FREE BOOK: http://www.cs.bham.ac.uk/research/cogaff/crp/
FREE TOOLS: http://www.cs.bham.ac.uk/research/poplog/packages/simagent.html
|