IMSL routines attempt to detect user errors and handle them in a way that provides as much information to the user as possible. To do this, we recognize various levels of severity of errors, and we also consider the extent of the error in the context of the purpose of the routine; a trivial error in one situation may be serious in another. IMSL routines attempt to report as many errors as they can reasonably detect. Multiple errors present a difficult problem in error detection because input is interpreted in an uncertain context after the first error is detected.

In some cases, the user’s input may be mathematically correct, but because of limitations of the computer arithmetic and of the algorithm used, it is not possible to compute an answer accurately. In this case, the assessed degree of accuracy determines the severity of the error. In cases where the routine computes several output quantities, if some are not computable but most are, an error condition exists. The severity depends on an assessment of the overall impact of the error.

If the user’s input is regarded as meaningless, such as N = –1 when “N” is the number of equations, the routine prints a message giving the value of the erroneous input argument(s) and the reason for the erroneous input. The routine will then cause the user’s program to stop. An error in which the user’s input is meaningless is the most severe error and is called a terminal error. Multiple terminal error messages may be printed from a single routine.

In many cases, the best way to respond to an error condition is simply to correct the input and rerun the program. In other cases, the user may want to take actions in the program itself based on errors that occur. An error that may be used as the basis for corrective action within the program is called an informational error. If an informational error occurs, a user‑retrievable code is set. A routine can return at most one informational error for a single reference to the routine. The codes for the informational error codes are printed in the error messages.

In addition to informational errors, IMSL routines issue error messages for which no user‑retrievable code is set. Multiple error messages for this kind of error may be printed. These errors, which generally are not described in the documentation, include terminal errors as well as less serious errors. Corrective action within the calling program is not possible for these errors.

Five levels of severity of errors are defined in the MATH/LIBRARY Special Functions. Each level has an associated PRINT attribute and a STOP attribute. These attributes have default settings (YES or NO), but they may also be set by the user. The purpose of having multiple error severity levels is to provide independent control of actions to be taken for errors of different severity. Upon return from an IMSL routine, exactly one error state exists. (A code 0 “error” is no informational error.) Even if more than one informational error occurs, only one message is printed (if the PRINT attribute is YES). Multiple errors for which no corrective action within the calling program is reasonable or necessary result in the printing of multiple messages (if the PRINT attribute for their severity level is YES). Errors of any of the severity levels except level 5 may be informational errors.

The user can set PRINT and STOP attributes by calling ERSET as described in “Routines for Error Handling.”

It is possible that a user’s program may call an IMSL routine that in turn calls a nested sequence of lower‑level IMSL routines. If an error occurs at a lower level in such a nest of routines and if the lower‑level routine cannot pass the information up to the original user‑called routine, then a traceback of the routines is produced. The only common situation in which this can occur is when an IMSL routine calls a user‑supplied routine that in turn calls another IMSL routine.

There are three ways in which the user may interact with the IMSL error handling system: (1) to change the default actions, (2) to retrieve the integer code of an informational error so as to take corrective action, and (3) to determine the severity level of an error. The routines to use are ERSET, IERCD and N1RTY, respectively.