sb-sprof self-test fails with FLOATING-POINT-INEXACT error on darwin

[Cyrus Harmon]

after building with lichtblau's recent changes (commit 2b29a7c2b236cfab1d4d06311e84414abba71b4c), I see the following when the build reaches the sb-sprof self test:

; SYS:CONTRIB;SB-SPROF;SB-SPROF.FASL.NEWEST written
; compilation finished in 0:00:00.938
* /Users/sly/projects/sbcl/sbcl/src/runtime/sbcl --noinform --core /Users/sly/projects/sbcl/sbcl/output/sbcl.core --lose-on-corruption --disable-debugger --no-sysinit --no-userinit --eval '(load (format nil "SYS:CONTRIB;~:@(~A~);TEST.LISP" "sb-sprof"))' </dev/null
unhandled FLOATING-POINT-INEXACT in thread #<SB-THREAD:THREAD
                                             "main thread" RUNNING
                                              {10029F9D03}>:
  arithmetic error FLOATING-POINT-INEXACT signalled

0: (SB-DEBUG::MAP-BACKTRACE
    #<CLOSURE (LAMBDA # :IN BACKTRACE) {1002C934FB}>
    :START
    0
    :COUNT
    128)
1: (BACKTRACE 128 #<SYNONYM-STREAM :SYMBOL SB-SYS:*STDERR* {1000174843}>)
2: (SB-DEBUG::DEBUGGER-DISABLED-HOOK
    #<FLOATING-POINT-INEXACT {1002C5DA13}>
    #<unavailable argument>)
3: (SB-DEBUG::RUN-HOOK
    *INVOKE-DEBUGGER-HOOK*
    #<FLOATING-POINT-INEXACT {1002C5DA13}>)
4: (INVOKE-DEBUGGER #<FLOATING-POINT-INEXACT {1002C5DA13}>)
5: (ERROR FLOATING-POINT-INEXACT)
6: (SB-VM:SIGFPE-HANDLER
    #<unavailable argument>
    #.(SB-SYS:INT-SAP #X011FE800)
    #<unavailable argument>)
7: ((FLET #:WITHOUT-INTERRUPTS-BODY-20 :IN SB-SYS:INVOKE-INTERRUPTION))
8: ((FLET SB-THREAD::EXEC :IN SB-SYS:INVOKE-INTERRUPTION))
9: ((FLET #:WITHOUT-INTERRUPTS-BODY-11 :IN SB-SYS:INVOKE-INTERRUPTION))
10: (SB-SYS:INVOKE-INTERRUPTION
     #<CLOSURE (FLET SB-UNIX::INTERRUPTION :IN SB-SYS:ENABLE-INTERRUPT)
       {11FE1EB}>)
11: (SB-SYS:INVOKE-INTERRUPTION
     #<CLOSURE (FLET SB-UNIX::INTERRUPTION :IN SB-SYS:ENABLE-INTERRUPT)
       {11FE1EB}>)[:EXTERNAL]
12: ((FLET SB-UNIX::RUN-HANDLER :IN SB-SYS:ENABLE-INTERRUPT)
     8
     #.(SB-SYS:INT-SAP #X011FE800)
     #.(SB-SYS:INT-SAP #X011FE868))
13: ("foreign function: call_into_lisp")
14: ("foreign function: funcall3")
15: ("foreign function: interrupt_handle_now")
16: ("foreign function: interrupt_handle_now_handler")
17: ("foreign function: _sigtramp")

to reproduce: make.sh

sbcl version: "1.1.2.35-2b29a7c"

(sly@orcia):~/projects/sbcl/sbcl$ uname -a
Darwin orcia 12.2.1 Darwin Kernel Version 12.2.1: Thu Oct 18 12:13:47 PDT 2012; root:xnu-2050.20.9~1/RELEASE_X86_64 x86_64

*features*

(:QUICKLISP :ASDF2 :ASDF :ASDF-UNICODE :ALIEN-CALLBACKS :ANSI-CL :BSD
 :C-STACK-IS-CONTROL-STACK :COMMON-LISP :COMPARE-AND-SWAP-VOPS
 :COMPLEX-FLOAT-VOPS :CYCLE-COUNTER :DARWIN :DARWIN9-OR-BETTER :FLOAT-EQL-VOPS
 :GENCGC :IEEE-FLOATING-POINT :INLINE-CONSTANTS :INODE64 :LINKAGE-TABLE
 :LITTLE-ENDIAN :MACH-EXCEPTION-HANDLER :MACH-O :MEMORY-BARRIER-VOPS
 :MULTIPLY-HIGH-VOPS :OS-PROVIDES-BLKSIZE-T :OS-PROVIDES-DLADDR
 :OS-PROVIDES-DLOPEN :OS-PROVIDES-PUTWC :OS-PROVIDES-SUSECONDS-T
 :RAW-INSTANCE-INIT-VOPS :SB-DOC :SB-EVAL :SB-LDB :SB-PACKAGE-LOCKS
 :SB-SOURCE-LOCATIONS :SB-TEST :SB-THREAD :SB-UNICODE :SBCL
 :STACK-ALLOCATABLE-CLOSURES :STACK-ALLOCATABLE-FIXED-OBJECTS
 :STACK-ALLOCATABLE-LISTS :STACK-ALLOCATABLE-VECTORS
 :STACK-GROWS-DOWNWARD-NOT-UPWARD :UD2-BREAKPOINTS :UNIX
 :UNWIND-TO-FRAME-AND-CALL-VOP :X86-64)

[David Lichteblau (david-lichteblau)]

First of all, mea culpa that I've pushed a test (and enabled it) as part of a different commit. The sprof stuff should have come first as a separate change.

As to the failure mode though:

If you copy&paste the (defun consing-test ...) into test.lisp, you should get a stand-alone test case that is basically independent of the SBCL version. Does that test.lisp, if compiled and loaded, also fail in an older SBCL on Darwin, e.g. 1.1.2?

[Paul Khuong (pvk)]

FWIW, we have FP trap masking issues on OS X: division by zeros seem to be reported as inexact on recent darwin/x86-64. So, we get values from <sys/signal.h>, but we seem to be supposed to use the constants in <fenv.h> instead (because, of course, they don't coïncide).

[Douglas Katzman (dougk)]

sb-sprof is tested by parallel-exec as well as its own tests, and doesn't fail.
So this is "fixed" though it would be more accurate to say that this bug is simply obsolete. I have no idea when sprof started working better.