2.3.11 broken on macOS 14 x86_64

Bug #2045745 reported by Kirill A. Korinskiy
10
This bug affects 2 people
Affects Status Importance Assigned to Milestone
SBCL
Invalid
Undecided
Unassigned

Bug Description

MacPorts build bots discovered a regression between 7e5c8161aeb33a11f6e853c40a7eedb6a6c38971 sbcl-2.3.11: https://github.com/sbcl/sbcl/compare/7e5c8161aeb33a11...sbcl-2.3.11

It fails as:

;;; Loading #P"/opt/local/var/macports/build/_opt_bblocal_var_buildworker_ports_build_ports_lang_sbcl/sbcl/work/sbcl-2.3.11/obj/from-host/src/compiler/float-tran.fas"

Condition of type: SIMPLE-PACKAGE-ERROR
Cannot intern symbol "MAKE-DOUBLE-FLOAT" in locked package #<"COMMON-LISP" package>.
Available restarts:

1. (CONTINUE) Ignore lock and proceed.
2. (RECOMPILE) Recompile
3. (ABORT-BUILD) Abort building SBCL.
4. (RESTART-TOPLEVEL) Go back to Top-Level REPL.

Broken at SB-COLD::LOAD-OR-CLOAD-XCOMPILER. In: #<process TOP-LEVEL 0x105240ee0>.
 File: #P"/opt/local/var/macports/build/_opt_bblocal_var_buildworker_ports_build_ports_lang_sbcl/sbcl/work/sbcl-2.3.11/src/cold/defun-load-or-cload-xcompiler.lisp" (Position #5133)
>>

Revision history for this message
Kirill A. Korinskiy (catap) wrote :
Revision history for this message
Kirill A. Korinskiy (catap) wrote :
Revision history for this message
Sergey Fedorov (barracuda156) wrote :

`sbcl` 2.3.11 also broken on Sonoma aarch64.

Revision history for this message
Sergey Fedorov (barracuda156) wrote :

sbcl build log from Sonoma / aarch64

Revision history for this message
Kirill A. Korinskiy (catap) wrote :

BTW I've tried to build 2.3.11 by the same ECL on macOS 12 and it works very well.

Revision history for this message
Douglas Katzman (dougk) wrote :

This is surely a problem in ECL, not SBCL.

The file 'float-tran.lisp' contains exactly two source lines with an explicit CL: package prefix, not counting the 1 comment line that has it. The symbols referred to are CL:LONG-FLOAT, CL:DOUBLE-FLOAT, and CL:SINGLE-FLOAT. So how pray tell does ECL believe that the SBCL code is trying to intern "MAKE-DOUBLE-FLOAT" in the COMMON-LISP package? There's no explicit call to INTERN, so it's likely a bug in the ECL loader (since it looks like COMPILE-FILE finished).
And regarding the complaints about DEFTRANSFORM which is SBCL's macro, why is it trying to say that the argument types are wrong?

I use CCL to compile SBCL, and that worked, so to suppose that both CCL and SBCL have "counteracting bugs" (i.e. a bug which allows the thing begin compiled to actually intern a new symbol in the CL package), you should really take this up with the ECL folks. Now that said, if you can identify a specific breaking change in SBCL, it's conceivable we might be able to find a workaround by expressing something differently.

Revision history for this message
Stas Boukarev (stassats) wrote :

Old ECL on github builds SBCL fine.

Changed in sbcl:
status: New → Incomplete
Revision history for this message
Kirill A. Korinskiy (catap) wrote :

Stas, the same version if ECL builds SBCL-2.3.10 on macOS 12 just fine. The issue is macOS 14.

BTW if you may distribute a binary which works on macOS 14, it allows to avoid using ECL to bootstrap SBCL.

Revision history for this message
Stas Boukarev (stassats) wrote :

It's not, I grabbed ecl from git and I get the same error on linux.

Revision history for this message
Kirill A. Korinskiy (catap) wrote :

I can reproduce the issue on macOS 13 with Xcode 15, and I can't reproduce this issue on different ECL.

So, this defently not SBCL issue

Revision history for this message
Kirill A. Korinskiy (catap) wrote :
Changed in sbcl:
status: Incomplete → Invalid
To post a comment you must log in.
This report contains Public information  
Everyone can see this information.

Other bug subscribers

Remote bug watches

Bug watches keep track of this bug in other bug trackers.