asksign-p-or-n (was RE: Regression in integration)

• Subject: asksign-p-or-n (was RE: Regression in integration)
• From: Barton Willis
• Date: Fri, 24 May 2013 12:10:26 +0000

The function asksign-p-or-n isn't an orphan, but running the testsuite, it is never called.  
The function simpexpt calls radmabs, and radmabs calls  asksign-p-or-n.  Apparently 
the call to radmabs is involved in the dubious simplification (x^a)^b -> abs(x)^(a*b). 
(See source code comment) Huh--what conditions are checked? 

  (%i9)   (x^a)^b /( abs(x)^(a*b));
  (%o9) (x^a)^b/abs(x)^(a*b)

  (%i10) psubst([a=7,b=3,x=-1],%);
  (%o10) -1

The function simpexpt is 343 lines with 23 calls to go :(

--Barton

• Prev by Date: Reading 32-bit IEEE floats from binary file ??
• Next by Date: Regression in integration
• Previous by thread: asksign bug: integrate(exp(a*x),x,0,inf);
• Next by thread: asksign(x^2) very broken! WAS problem with 'is zero or nonzero'
• Index(es):
  • Date
  • Thread