The Ada Reference Manual defines the concept of invalid values (see RM 13.9.1). The primary source of invalid values is uninitialized variables. A scalar variable that is left uninitialized may contain an invalid value; the concept of invalid does not apply to access or composite types.
It is an error to read an invalid value, but the RM does not require
run-time checks to detect such errors, except for some minimal
checking to prevent erroneous execution (i.e. unpredictable
behavior). This corresponds to the -gnatVd switch below,
which is the default. For example, by default, if the expression of a
case statement is invalid, it will raise Constraint_Error rather than
causing a wild jump, and if an array index on the left-hand side of an
assignment is invalid, it will raise Constraint_Error rather than
overwriting an arbitrary memory location.
The -gnatVa may be used to enable additional validity checks,
which are not required by the RM. These checks are often very
expensive (which is why the RM does not require them). These checks
are useful in tracking down uninitialized variables, but they are
not usually recommended for production builds, and in particular
we do not recommend using these extra validity checking options in
combination with optimization, since this can confuse the optimizer.
If performance is a consideration, leading to the need to optimize,
then the validity checking options should not be used.
The other -gnatV`x' switches below allow finer-grained
control; you can enable whichever validity checks you desire. However,
for most debugging purposes, -gnatVa is sufficient, and the
default -gnatVd (i.e. standard Ada behavior) is usually
sufficient for non-debugging use.
The -gnatB switch tells the compiler to assume that all
values are valid (that is, within their declared subtype range)
except in the context of a use of the Valid attribute. This means
the compiler can generate more efficient code, since the range
of values is better known at compile time. However, an uninitialized
variable can cause wild jumps and memory corruption in this mode.
The -gnatV`x' switch allows control over the validity
checking mode as described below.
The x argument is a string of letters that
indicate validity checks that are performed or not performed in addition
to the default checks required by Ada as described above.
-gnatVa`All validity checks.'
All validity checks are turned on.
That is, -gnatVa is
equivalent to gnatVcdfimoprst.
-gnatVc`Validity checks for copies.'
The right hand side of assignments, and the initializing values of object declarations are validity checked.
-gnatVd`Default (RM) validity checks.'
Some validity checks are done by default following normal Ada semantics
(RM 13.9.1 (9-11)).
A check is done in case statements that the expression is within the range
of the subtype. If it is not, Constraint_Error is raised.
For assignments to array components, a check is done that the expression used
as index is within the range. If it is not, Constraint_Error is raised.
Both these validity checks may be turned off using switch -gnatVD.
They are turned on by default. If -gnatVD is specified, a subsequent
switch -gnatVd will leave the checks turned on.
Switch -gnatVD should be used only if you are sure that all such
expressions have valid values. If you use this switch and invalid values
are present, then the program is erroneous, and wild jumps or memory
overwriting may occur.
-gnatVe`Validity checks for elementary components.'
In the absence of this switch, assignments to record or array components are
not validity checked, even if validity checks for assignments generally
(-gnatVc) are turned on. In Ada, assignment of composite values do not
require valid data, but assignment of individual components does. So for
example, there is a difference between copying the elements of an array with a
slice assignment, compared to assigning element by element in a loop. This
switch allows you to turn off validity checking for components, even when they
are assigned component by component.
-gnatVf`Validity checks for floating-point values.'
In the absence of this switch, validity checking occurs only for discrete
values. If -gnatVf is specified, then validity checking also applies
for floating-point values, and NaNs and infinities are considered invalid,
as well as out of range values for constrained types. Note that this means
that standard IEEE infinity mode is not allowed. The exact contexts
in which floating-point values are checked depends on the setting of other
options. For example, -gnatVif or -gnatVfi
(the order does not matter) specifies that floating-point parameters of mode
in should be validity checked.
-gnatVi`Validity checks for ‘‘in‘‘ mode parameters.'
Arguments for parameters of mode in are validity checked in function
and procedure calls at the point of call.
-gnatVm`Validity checks for ‘‘in out‘‘ mode parameters.'
Arguments for parameters of mode in out are validity checked in
procedure calls at the point of call. The 'm' here stands for
modify, since this concerns parameters that can be modified by the call.
Note that there is no specific option to test out parameters,
but any reference within the subprogram will be tested in the usual
manner, and if an invalid value is copied back, any reference to it
will be subject to validity checking.
-gnatVn`No validity checks.'
This switch turns off all validity checking, including the default checking
for case statements and left hand side subscripts. Note that the use of
the switch -gnatp suppresses all run-time checks, including
validity checks, and thus implies -gnatVn. When this switch
is used, it cancels any other -gnatV previously issued.
-gnatVo`Validity checks for operator and attribute operands.'
Arguments for predefined operators and attributes are validity checked.
This includes all operators in package Standard,
the shift operators defined as intrinsic in package Interfaces
and operands for attributes such as Pos. Checks are also made
on individual component values for composite comparisons, and on the
expressions in type conversions and qualified expressions. Checks are
also made on explicit ranges using .. (e.g., slices, loops etc).
-gnatVp`Validity checks for parameters.'
This controls the treatment of parameters within a subprogram (as opposed
to -gnatVi and -gnatVm which control validity testing
of parameters on a call. If either of these call options is used, then
normally an assumption is made within a subprogram that the input arguments
have been validity checking at the point of call, and do not need checking
again within a subprogram). If -gnatVp is set, then this assumption
is not made, and parameters are not assumed to be valid, so their validity
will be checked (or rechecked) within the subprogram.
-gnatVr`Validity checks for function returns.'
The expression in return statements in functions is validity
checked.
-gnatVs`Validity checks for subscripts.'
All subscripts expressions are checked for validity, whether they appear on the right side or left side (in default mode only left side subscripts are validity checked).
-gnatVt`Validity checks for tests.'
Expressions used as conditions in if, while or exit
statements are checked, as well as guard expressions in entry calls.
The -gnatV switch may be followed by a string of letters
to turn on a series of validity checking options.
For example, -gnatVcr
specifies that in addition to the default validity checking, copies and
function return expressions are to be validity checked.
In order to make it easier to specify the desired combination of effects,
the upper case letters CDFIMORST may
be used to turn off the corresponding lower case option.
Thus -gnatVaM turns on all validity checking options except for
checking of in out parameters.
The specification of additional validity checking generates extra code (and
in the case of -gnatVa the code expansion can be substantial).
However, these additional checks can be very useful in detecting
uninitialized variables, incorrect use of unchecked conversion, and other
errors leading to invalid values. The use of pragma Initialize_Scalars
is useful in conjunction with the extra validity checking, since this
ensures that wherever possible uninitialized variables have invalid values.
See also the pragma Validity_Checks which allows modification of
the validity checking mode at the program source level, and also allows for
temporary disabling of validity checks.