sbcl-devel

[Sbcl-devel] SBCL on FreeBSD

[Raymond A. W. <rw...@c2...>]

	Yesterday I started working on SBCL under FreeBSD again
(currently 3.4-STABLE), and eventually got a running image (I had to
perform the last step in make.sh by hand, as I had to change the
definitions for "struct stat" and "extract-stat-results", and
reevaluate the definitions for stat, lstat and fstat (as they use the
macro extract-stat-results).

	I'm now running make.sh again, with my changes merged into
unix.lisp. I cannot claim that I got the merge right on my first try,
but I think that the version I'm compiling just now will work.

	How do I go about getting my work merged into sbcl (actually,
it's not a lot of changed code, despite the fact that I've been
working on this for a while. Make of that what you will :-)

	Would it be an idea if I prepared a diff against 0.6.3, plus
the files I've imported/updated from CMUCL (FreeBSD-os.[hc] and
Config.FreeBSD_gencgc)? This way, somebody else with a FreeBSD machine
could verify the integrity of the patch. It's also an idea to build
for Linux with my patches present, of course...

	//Raymond.    

Re: [Sbcl-devel] SBCL on FreeBSD

[William H. N. <wil...@ai...>]

I'm sorry about the late reply. I've been moving to a new apartment,
and my computers moved out of the old apartment with working phone
connection this weekend, and only got a new working phone connection
today.

On Mon, May 01, 2000 at 07:31:50PM +0200, Raymond A. Wiker wrote:
> 	Yesterday I started working on SBCL under FreeBSD again
> (currently 3.4-STABLE), and eventually got a running image (I had to
> perform the last step in make.sh by hand, as I had to change the
> definitions for "struct stat" and "extract-stat-results", and
> reevaluate the definitions for stat, lstat and fstat (as they use the
> macro extract-stat-results).
> 
> 	I'm now running make.sh again, with my changes merged into
> unix.lisp. I cannot claim that I got the merge right on my first try,
> but I think that the version I'm compiling just now will work.

This is great!
 
> 	How do I go about getting my work merged into sbcl (actually,
> it's not a lot of changed code, despite the fact that I've been
> working on this for a while. Make of that what you will :-)
>
> 	Would it be an idea if I prepared a diff against 0.6.3, plus
> the files I've imported/updated from CMUCL (FreeBSD-os.[hc] and
> Config.FreeBSD_gencgc)? This way, somebody else with a FreeBSD machine
> could verify the integrity of the patch. It's also an idea to build
> for Linux with my patches present, of course...

That sounds like a good approach. I don't have any real experience
with merging branches and patches, but it's probably time to learn,
and you make it sound like a manageable amount of change. If you
can test for Linux with your patches present before you send them
on to me, that could make my work easier, but if that's inconvenient,
just send the patches and I'll try to sort them out.

(I have a FreeBSD machine, so I'll be able to test the results.)

-- 
William Harold Newman <wil...@ai...>
software consultant
PGP key fingerprint 85 CE 1C BA 79 8D 51 8C  B9 25 FB EE E0 C3 E5 7C

Re: [Sbcl-devel] SBCL on FreeBSD

[Raymond W. <ra...@pc...>]

William Harold Newman writes:
 > On Mon, May 01, 2000 at 07:31:50PM +0200, Raymond A. Wiker wrote:
 > > 	I'm now running make.sh again, with my changes merged into
 > > unix.lisp. I cannot claim that I got the merge right on my first try,
 > > but I think that the version I'm compiling just now will work.
 > 
 > This is great!

	Quick note: I had to run clean.sh to get rid of some
(generated) files under src/assembly/target before running make.sh. 

	I also had to add ":ignore-failure-p" to the entry for
code/seq in stems-and-flags.lisp-expr, as there was a warning about
"unused variable DOC" when compiling this file (using CMUCL).

	With the modified unix.lisp I got a running sbcl+sbcl.core,
which is working well enough to compile sbcl itself. I'll prepare a
patch later; I can also put a binary distribution for FreeBSD on a web
server I have access to, if there's interest. This binary distribution
works for both FreeBSD 3.4 and 4.0, BTW.

	I'm currently working on getting sbcl to compile under FreeBSD
4.0. There has been a change to struct sigcontext lately, and Tim
Moore (I think) submitted patches to the CMUCL mailing list that I'm
trying to integrate. Should I put out a set of patches for FreeBSD 3.4
now, or should I wait until I have the 4.0 stuff in place? It's not a
problem for me to provide two sets of patches, with the first being
3.4-only, and the second being for both 3.4 and 4.0.

 > That sounds like a good approach. I don't have any real experience
 > with merging branches and patches, but it's probably time to learn,
 > and you make it sound like a manageable amount of change. If you
 > can test for Linux with your patches present before you send them
 > on to me, that could make my work easier, but if that's inconvenient,
 > just send the patches and I'll try to sort them out.

	Integrating my changes should be straightforward - my changes
relative to 0.6.0 merged into 0.6.3 with only a couple of instances
where manual intervention was required.

	I'll see if I can get a friend with a Linux machine to
test-compile SBCL-0.6.3 with my patches.

	//Raymond.

Re: [Sbcl-devel] SBCL on FreeBSD

[William H. N. <wil...@ai...>]

On Sat, May 06, 2000 at 02:16:01PM +0200, Raymond Wiker wrote:
> 	I'm currently working on getting sbcl to compile under FreeBSD
> 4.0. There has been a change to struct sigcontext lately, and Tim
> Moore (I think) submitted patches to the CMUCL mailing list that I'm
> trying to integrate. Should I put out a set of patches for FreeBSD 3.4
> now, or should I wait until I have the 4.0 stuff in place? It's not a
> problem for me to provide two sets of patches, with the first being
> 3.4-only, and the second being for both 3.4 and 4.0.

[..]

> 	Integrating my changes should be straightforward - my changes
> relative to 0.6.0 merged into 0.6.3 with only a couple of instances
> where manual intervention was required.
> 
> 	I'll see if I can get a friend with a Linux machine to
> test-compile SBCL-0.6.3 with my patches.

I'm actually still running 3.3 on my FreeBSD machine, but I know I
should upgrade to 4.0 one of these days, and this sounds like a good
enough reason.

If you think the patches you have for 3.4 will work on 3.3, why don't
you just send them to me now -- testing on your friend's Linux machine
probably isn't that important, I can do that myself -- and I'll try to
integrate them into the 0.6.4 release.

If you don't think the patches for 3.4 will work on 3.3, then I'll
probably wait until you have patches for 4.0, at which point I'll
finally upgrade my FreeBSD to 4.0 and test that way.
 
-- 
William Harold Newman <wil...@ai...>
software consultant
PGP key fingerprint 85 CE 1C BA 79 8D 51 8C  B9 25 FB EE E0 C3 E5 7C

Re[2]: [Sbcl-devel] SBCL on FreeBSD

[Raymond W. <rw...@c2...>]

Saturday, May 06, 2000, 17:33:29, you wrote:

WHN> I'm actually still running 3.3 on my FreeBSD machine, but I know I
WHN> should upgrade to 4.0 one of these days, and this sounds like a good
WHN> enough reason.

    You may want to hold on for a bit. I discovered that my ISDN card
is not supported under FreeBSD 4.0, which means that I have to either
reboot my home machine with Win95 (yuk) to receive/send mail, or go
through my work machine. Actually, my work machine is a laptop running
FreeBSD 4.0, so it's not a big problem for me to prepare a set of
patches at home, copy them over to the laptop and send them from work.

    I *think* my patches should work under 3.3 as well as 3.4, but I
haven't tried this.

    The current state of affairs is that I can build SBCL under
FreeBSD, and use the newly-built sbcl and sbcl.core to build SBCL
again. The second build process fails when sbcl tries to write
output/cold-sbcl.core; I expect that this is because of a conflict
between the running SBCL and object files for the "new" SBCL. My guess
is that this does not happen when using CMUCL because CMUCL uses
different package names.

    I suspect that I may have misunderstood something about when to
use #!+/- instead of #+/-; particularly in src/code/unix.lisp. I had
tests of the form #+FreeBSD here, which I replaced with #!+FreeBSD
this morning. It didn't help; I'm still getting an error message akin
to "Open failed: no such file or directory". I suspect that unix-stat
is getting munged somehow when unix.lisp-obj is getting loaded - does
this sound reasonable?

WHN> If you think the patches you have for 3.4 will work on 3.3, why don't
WHN> you just send them to me now -- testing on your friend's Linux machine
WHN> probably isn't that important, I can do that myself -- and I'll try to
WHN> integrate them into the 0.6.4 release.

    I'll try to send the patches today or tomorrow. This will be a
set of patches that should work for both 3.4 and 4.0. If this doesn't
work for you, I could try and make a set for 3.4 only - the main
difference is related to a recent change in struct sigcontext.

WHN> If you don't think the patches for 3.4 will work on 3.3, then I'll
WHN> probably wait until you have patches for 4.0, at which point I'll
WHN> finally upgrade my FreeBSD to 4.0 and test that way.

-- 
Best regards,
 Raymond Wiker                            mailto:rw...@c2...