Error checking
Almost every system call or library routine can,
somehow,
encounter an error during its operation.
While many of these occur only because of
programmer error,
each such problem indicates
a failure of either the system,
the calling program or a transient parameter
like access permission or available memory.
If a programmer chooses to ignore a
reported error,
the result is a command that,
should some basic assumption of the system fail,
could corrupt its environment.
For trusted commands,
therefore,
every possible error return must be
checked and reported.
This rule is not always followed to the letter,
since in some cases it is
more efficient to detect the error
case downstream from the actual failure.
Ignoring errors is risky and
should not be done without strong justification.
Next topic:
Signal handling
Previous topic:
Privilege and special access in shared private routines
© 2004 The SCO Group, Inc. All rights reserved.
UnixWare 7 Release 7.1.4 - 27 April 2004