[cmucl-imp] Upcoming release

[Raymond Toy]

On 10/25/11 2:49 PM, Robert Swindells wrote:
> Raymond Toy wrote:
>> On 10/24/11 3:20 PM, Robert Swindells wrote:
> .> Raymond Toy wrote:
>>>> I noticed that it has been a year since our last release, so it's time
>>>> once again for another release.  A new branch has been created
>>>> (RELEASE-20c-BRANCH) for the release.
>>>>
>>>> This time, I'm not going to do any pre-release.  We'll make just the one
>>>> build due at the end of this month for the 20c release.  If it's broken
>>>> for some reason, we'll send a patch if possible.  If it's really broken,
>>>> we'll recommend people to use the following snapshot.  (Of course,
>>>> because of the release, there will not be a snapshot for Nov.)
>>> I think that for NetBSD the recommendation will have to be to stick
>>> with the 2011-10 snapshot. I can rebuild this for later OS releases.
>>>
>>> The build scripts don't work anymore.
>> Bummer.  In what way do the build scripts not work anymore?  Did you
>> remember to use the boot-20c bootstrap file?  There were really very few
>> changes between 2011-10 and the release candidate.  Other than changing
>> the fasl file version, I think the only possible issue would be the
>> denormal trap handling.
> With create-target.sh:
>
> % src/tools/create-target.sh build-2 x86_netbsd NetBSD
> (create-target.sh) Settings: bld_dir=build-2 lisp_variant=x86_netbsd motif_variant=NetBSD
> cd: can't cd to motif/server
> (create-target.sh) not OK: Can't cd motif/server

Ok.  Now that I know how things are failing, I'll take a look at this. 
Alex changed some of the scripts around a little recently, so that
probably explains the failure.  Should be simple to fix.  It will
definitely be fixed before the release.

Thanks for testing this *before* the release label was set. :-)

Ray

>
>
> I'm fairly sure that I have used build-all.sh in the past but I think
> it is just getting confused if you don't use specify the options.
>
> Creating the directories beforehand has this problem:
>
> % src/tools/build-all.sh -o "unicode/lisp" -8 "non-unicode/lisp"
> + src/tools/build.sh -f x87 -b build-8bit  -C  -P -o non-unicode/lisp
> //starting build: Tue Oct 25 19:41:40 BST 2011
> cmulisp: not found
>         0.01 real         0.00 user         0.00 sys
> Failed: src/tools/build-world.sh
> + src/tools/build.sh -f sse2 -b build-8bit  -C  -P -o non-unicode/lisp
> //starting build: Tue Oct 25 19:41:41 BST 2011
> cmulisp: not found
>         0.01 real         0.00 user         0.00 sys
> Failed: src/tools/build-world.sh
> + set +x
> + src/tools/build.sh -f x87 -b build  -C  -P -o unicode/lisp
> //starting build: Tue Oct 25 19:41:42 BST 2011
> cmulisp: not found
>         0.01 real         0.00 user         0.00 sys
> Failed: src/tools/build-world.sh
> + src/tools/build.sh -f sse2 -b build  -C  -P -o unicode/lisp
> //starting build: Tue Oct 25 19:41:42 BST 2011
> cmulisp: not found
>         0.01 real         0.00 user         0.00 sys
> Failed: src/tools/build-world.sh
> + set +x
>
> Trying to get build-all.sh to create the directories doesn't work for
> me either, I get:
>
> % src/tools/build-all.sh -o unicode/lisp -8 non-unicode/lisp -C "x86_netbsd NetBSD"
> + src/tools/build.sh -f x87 -b build-8bit  -C x86_netbsd NetBSD -P -o non-unicode/lisp
> //starting build: Tue Oct 25 22:47:12 BST 2011
> build-8bit-2 isn't a directory
> Failed: src/tools/clean-target.sh
> + src/tools/build.sh -f sse2 -b build-8bit  -C x86_netbsd NetBSD -P -o non-unicode/lisp
> //starting build: Tue Oct 25 22:47:13 BST 2011
> build-8bit-2 isn't a directory
> Failed: src/tools/clean-target.sh
> + set +x
> + src/tools/build.sh -f x87 -b build  -C x86_netbsd NetBSD -P -o unicode/lisp
> //starting build: Tue Oct 25 22:47:13 BST 2011
> build-2 isn't a directory
> Failed: src/tools/clean-target.sh
> + src/tools/build.sh -f sse2 -b build  -C x86_netbsd NetBSD -P -o unicode/lisp
> //starting build: Tue Oct 25 22:47:13 BST 2011
> build-2 isn't a directory
> Failed: src/tools/clean-target.sh
> + set +x
>
> The problem with stage-1 crashing was just me being stupid and building
> it on the wrong operating system version, time_t is still 32 bit on the
> released versions of NetBSD and the CMUCL source matches this.
>
> It would be helpful to get the scripts fixed though.
>
> Robert Swindells
>