From: Albert_Rich@msn.com   
      
   On Friday, July 14, 2017 at 4:01:31 PM UTC-10, Nasser M. Abbasi wrote:   
   > On 7/13/2017 5:56 PM, Nasser M. Abbasi wrote:   
   >    
   > >    
   > > Relatively, the systems did better in last build with the 58,469   
   > > integrals   
   >    
   > Ok, I found out why. It is my mistake. I re-wrote the whole build   
   > script again from scratch when adding sympy, and forgot that before   
   > I was counting a test which failed but has no known anti-derivative   
   > as PASSED.   
   >    
   >  From earlier build it says   
   >    
   > "Not solving a problem with no known antiderivative is counted as passed   
   test"   
   >    
   > But for new build, the same integral counts as failed now as I am not   
   checking   
   > if it has known anti-derivative or not.   
   >    
   > This explain the large difference in final score result. I should   
   > correct this and go back to the old counting method.   
   >    
   > --Nasser   
      
   Hello Nasser,   
      
   Glad you figured out the reason for the big increase in the number of failed   
   test results.   
      
   Certainly if no closed-form antiderivative exists, a system should be given a   
   passing grade of A for returning the integral unevaluated.  Note that problems   
   having no valid antiderivative are included in the test suite precisely   
   because such problems    
   frequently uncover infinite recursion bugs that result in stack exhaustion   
   messages and system hangs.   
      
   Sorry you have to run the entire test suite again. :(  Note that using a   
   random number generator to select say 10% of the 63,000+ test suite problems,   
   I could send you a reduced-size test suite.  This would make it easier for you   
   to run the test suite    
   more often, and small enough to test the slower systems you mentioned.     
      
   Another alternative would be for your test program to skip problems whose   
   integrand and/or optimal antiderivative involves special or hypergeometric   
   functions.  This “elementary-function-only” test option would allow for   
   fairly testing computer    
   algebra systems that do not support higher level functions, and be of more   
   interest to users not familiar with such functions.   
      
   Albert   
      
   --- SoupGate-Win32 v1.05   
    * Origin: you cannot sedate... all the things you hate (1:229/2)