ccc@meccts.UUCP (Chad C. Cooper) (08/27/86)
I have encountering a problem, which I believe is occuring in Borland's Turbo Database Toolbox, and I am interested in seeing if anybody else has had simular problems. We are using Turbo Pascal to develope a data base package. Borland, out of the goodness of their heart(and greed in their pocketbook), has developed routines to manage B-Trees. Great! This was the solution to us having to spend hours developing our own routines, so we would just have to write the editor, statistics, and printing routines. The symptoms of the problem are this: The data base is created. After openning the file and entering data, during the closing process it encounters an error 255 because it can't find the file to close. What happens on the disk, is a file that used to be 'STUDNT02.DAT' has been changed to 'STU^NT02.D^^' with the ^ being ASCII zero characters. The problem seems to be when the tree in memory filled. It then has to write the tree out to the disk and get a smaller section into memory. At this point, it seems to step on a buffer in memory that contains the directory entry for the current file that is open. At some point in time, during the close process(flushing both the Turbo and Dos related buffers) it rewrites that buffer onto the disk. We have attempted modifying the constants that define the B-Tree buffering and manipulation. We have made it more difficult to duplicate the bug, but have not been able to find a configuration that eliminates it. Has anyone else experienced anything like this? Is anyone out in netland from Borlund? Thanx for any responce. -c^3 "Just ignore any abnormal statements, it is just a flux in reality." - me Chad C. Cooper MECC Technical Services (612) 481-3686 ihnp4!dicome!meccts!ccc