DIVISION-BY-ZERO following (import-module "numpy") (V 2.2.1)

[mirkov]

Hello,

This is on MacBook Air M2, SBCL 2.5.0

After I issue an (import-module "numpy") I am dropped into the debugger with the "DIVISION-BY-ZERO" error

I suspect this is because I am on numpy V2.2.1 (which is not yet supported as per the documentation).

I am posting to confirm that this is the expected behavior at this time, and not me doing something incorrect. I installed this library only yesterday and these are my first steps in it (as in Python).

Looking forward to the numpy V2 support :-)

Full debug message:

  arithmetic error DIVISION-BY-ZERO signalled
     [Condition of type DIVISION-BY-ZERO]

  Restarts:
   0: [RETRY] Retry SLY mREPL evaluation request.
   1: [*ABORT] Return to SLY's top level.
   2: [ABORT] abort thread (#<THREAD tid=6147 "sly-channel-1-mrepl-remote-1" RUNNING {7005724033}>)

  Backtrace:
   0: ((FLET SB-UNIX::RUN-HANDLER :IN SB-UNIX::%INSTALL-HANDLER) 8 #.(SB-SYS:INT-SAP #X104FFF0C0) #.(SB-SYS:INT-SAP #X104FFF128))
   1: ("bogus stack frame")
   2: (IMPORT-MODULE "numpy" :AS NIL)
   3: (SB-INT:SIMPLE-EVAL-IN-LEXENV (IMPORT-MODULE "numpy") #<NULL-LEXENV>)
   4: (EVAL (IMPORT-MODULE "numpy"))
   5: ((LAMBDA NIL :IN SLYNK-MREPL::MREPL-EVAL-1))

[digikar99]

The division by zero error occurs even with numpy 1 :/.

As a temporary solution, you can use (float-features:with-float-traps-masked t (import-module "numpy")).

As a permanent solution, we need to look into how the float traps are handled in numpy vs lisp and potentially wrap module imports in float-features:with-float-traps-masked.