Request for merging "lisp-team" branch

The main update of the branch is sbcl 2.4.0.

The CI built it fine for x86-64 and i686, but has not tried yet on

other architectures (after several days jobs are still in "scheduled"

state).

If you can, please test the branch and report any issue here.

On Mon, Jan 08, 2024 at 01:06:30PM +0000, Guillaume Le Vaillant wrote:

I have sbcl@2.4.0 from the lisp-team branch failing to build on

powerpc64le and succeeding on aarch64-linux.

riscv64-linux failed for me too:

"obj/from-xc/src/code/early-float.lisp-obj" fatal error encountered in SBCL pid 553 tid 553:

internal error too early in init, can't recover

Error opening /dev/tty: No such device or address

Internal error #1 "An attempt was made to use an undefined FDEFINITION." at 0x4101560

SC: 0, Offset: 28 $1= 0x4f286dff: other pointer

Welcome to LDB, a low-level debugger for the Lisp runtime environment.

ldb>

--

Efraim Flashner <efraim@flashner.co.il> ????? ?????

GPG key = A28B F40C 3E55 1372 662D 14F7 41AA E7DC CA3D 8351

Confidentiality cannot be guaranteed on emails sent or received unencrypted

On Monday, January 8th, 2024 at 1:12 PM, Efraim Flashner <efraim@flashner.co.il> wrote:

I don't have a ppc64le machine, and I don't know about riscv64,

but SBCL 2.4.0 needs a patch to build on ppc64 [1].

[1]: https://github.com/sbcl/sbcl/commit/255f3ead060129aa097b62f10d054cdc4997a431

"Paul A. Patience" <paul@apatience.com> skribis:

I adapted the patch from upstream and added it to the lisp-team branch.

I tried testing the build with "guix build -s powerpc46le-linux sbcl"

on my x86-64 machine, but it failed indicating that the clisp used to

boostrap sbcl isn't working properly. I don't know if the problem comes

from qemu or if clisp is really broken on powerpc64le.

Efraim, could you check on your powerpc64le machine?

I also tried "guix build -s riscv64-linux sbcl", but it didn't work,

because some dependencies fail to build (like gnutls).

On Tue, Jan 09, 2024 at 02:10:17PM +0000, Guillaume Le Vaillant wrote:

Sure, I have it building on berlin now for ppc64le and on one of my

machines for riscv64. I'll let you know how they go.

--

Efraim Flashner <efraim@flashner.co.il> ????? ?????

GPG key = A28B F40C 3E55 1372 662D 14F7 41AA E7DC CA3D 8351

Confidentiality cannot be guaranteed on emails sent or received unencrypted

On Tue, Jan 09, 2024 at 02:10:17PM +0000, Guillaume Le Vaillant wrote:

Build was successful on ppc64le

--

Efraim Flashner <efraim@flashner.co.il> ????? ?????

GPG key = A28B F40C 3E55 1372 662D 14F7 41AA E7DC CA3D 8351

Confidentiality cannot be guaranteed on emails sent or received unencrypted

On Tue, Jan 09, 2024 at 02:10:17PM +0000, Guillaume Le Vaillant wrote:

riscv64-linux failure looks similar at first look to the ppc64le failure

"obj/from-xc/src/code/cold-error.lisp-obj"

"obj/from-xc/src/code/early-full-eval.lisp-obj"

"obj/from-xc/src/code/debug-var-io.lisp-obj"

"obj/from-xc/src/code/early-float.lisp-obj" fatal error encountered in SBCL pid 553 tid 553:

internal error too early in init, can't recover

Error opening /dev/tty: No such device or address

Internal error #1 "An attempt was made to use an undefined FDEFINITION." at 0x4101560

SC: 0, Offset: 28 $1= 0x4f286dff: other pointer

Welcome to LDB, a low-level debugger for the Lisp runtime environment.

ldb>

real 0m0.325s

user 0m0.066s

sys 0m0.242s

error: in phase 'build': uncaught exception:

%exception #<&invoke-error program: "sh" arguments: ("make.sh" "clisp" "--prefix=/gnu/store/wwy4kc4ary627b7igx5cxjsgkl8nq0yv-sbcl-2.4.0" "--dynamic-space-size=3072" "--with-sb-core-compression" "--with-sb-xref-for-internals" "--without-sb-simd") exit-status: 1 term-signal: #f stop-signal: #f>

phase `build' failed after 10113.1 seconds

--

Efraim Flashner <efraim@flashner.co.il> ????? ?????

GPG key = A28B F40C 3E55 1372 662D 14F7 41AA E7DC CA3D 8351

Confidentiality cannot be guaranteed on emails sent or received unencrypted

Efraim Flashner <efraim@flashner.co.il> skribis:

I reported the issue on riscv64 upstream

(https://bugs.launchpad.net/sbcl/+bug/2048869).

In case there is no fix for riscv64 right now, do you think we should

block sbcl updates, or should we update it anyway as it works on other

architectures, and hope that riscv64 will work again in the future?

On Wed, Jan 10, 2024 at 10:45:09AM +0000, Guillaume Le Vaillant wrote:

No need to wait for riscv64 for sbcl.

--

Efraim Flashner <efraim@flashner.co.il> ????? ?????

GPG key = A28B F40C 3E55 1372 662D 14F7 41AA E7DC CA3D 8351

Confidentiality cannot be guaranteed on emails sent or received unencrypted

Efraim Flashner <efraim@flashner.co.il> skribis:

Ok. I merged master in lisp-team, and if there are no new breakages,

I'll merge lisp-team in master.

Guillaume Le Vaillant <glv@posteo.net> skribis:

I just added a patch from upstream that should fix the riscv build.

now riscv build success for me

??? <zhengjunjie@yandex.com> skribis:

Great. Thanks for testing.

I think I'll merge lisp-team into master in a few hours.

lisp-team branch merged into master as

24f000a49a00b8dd3e9f51e08f0ef5567edf04aa.