[Date Index] [Thread Index] [Author Index]
[This is the response by Wolfram Research to the two previous messages. -- Steve Christensen.] -------------------------------------------------------------------------------- A recent posting to the mathgroup by Dr. Stefan Mueller raises some valid questions about bugs in Mathematica. As the manager of the Software Quality Assurance Department at Wolfram Research, I would like to address these issues. There was indeed a bug, as described by Dr. Mueller, in early versions of Mathematica 1.2 on all platforms. The bug was actually in the Together function, and was due to a fairly subtle logic error. This bug caused Together to give incorrect answers for expressions involving both integer and rational powers of the same subexpression. A very simple example of this problem: In:= Together[(a + b x)/(1 + Sqrt[a + b x])] Sqrt[a + b x] Out= ----------------- 1 + Sqrt[a + b x] ... which of course should give: a + b x Out= ----------------- 1 + Sqrt[a + b x] ... which is what the current versions of Mathematica do indeed give. Because Simplify calls Together (as do many other functions in our algebraic simplifi- cation system), this bug was manifested in Simplify as well, as shown by Dr. Mueller's example. As soon as this bug was discovered, a fix was made and a free upgrade was sent to all of our registered users. We sent out several thousand "fix" disks to our Macintosh customers. This was the first time in the history of Mathematica that Wolfram Research has had to send out upgrades of this kind, as this was the first time that a bug of this severity was uncovered in a released version of the program. Unfortunately, as Dr. Mueller's case demonstrates, some of our users did not receive their upgrade disks. We are currently investigating how and why this happened, and are taking steps to ensure that it does not happen again should such an occasion arise in the future. The policy of Wolfram Research is that if a bug is found that causes a function to give incorrect answers, and it is in a function used by most or all of our users, we will send out free upgrades to all of our registered users. If the bug is in a more specialized function, which does not affect other functions, free upgrades will be sent to registered users who request it. The letters accompanying any future such upgrades will explicitly state which functions are affected by the upgrade. Note, however, the importance of registering your copy of Mathematica: We cannot notify you of upgrades or send you bug fixes if your copy is not registered. More importantly, we are taking significant steps to ensure that such bugs do not appear in released versions of Mathematica. We have a very ambitious program for testing Mathematica, and we are always interested in hearing suggestions for further tests. For example, in recent months we have been comparing the output of Mathematica's special functions with the tables in Abramowitz and Stegun. (We found a couple of typographical errors in Abramowitz and Stegun; we didn't find any new bugs in Mathematica. I won't say anything about old bugs! ;-) ) We have a very intensive testing effort scheduled for this summer; the SQA department has hired seven additional summer interns to test various aspects of the program. Of course, testing can only find bugs; it cannot prove that there are no bugs. Although we try to keep the program bug-free, in a system of the size and complexity of Mathematica, some errors are inevitable. If you do find a bug in Mathematica, please send us a bug report by email to bugs at wri.com, by fax at (217) 398-0747 (to the attention of the SQA Department), or by snail mail to Software Quality Assurance, Wolfram Research, Inc., P.O. Box 6059, Champaign, IL 61826-6059. We can't fix them if we don't know about them! Your bug report will be acknowledged and you will be given a bug-tracking number which you may use if you wish to inquire about the status of the bug. If you have any general concerns or comments about the quality of Mathematica, please contact me by email at sqa at wri.com, or by fax or mail as above. Kate Ebneter Software Quality Assurance Manager Wolfram Research, Inc.