Next: GDB/MI Data Manipulation, Previous: GDB/MI Stack Manipulation, Up: GDB/MI
Variable objects are "object-oriented" MI interface for examining and changing values of expressions. Unlike some other MI interfaces that work with expressions, variable objects are specifically designed for simple and efficient presentation in the frontend. A variable object is identified by string name. When a variable object is created, the frontend specifies the expression for that variable object. The expression can be a simple variable, or it can be an arbitrary complex expression, and can even involve CPU registers. After creating a variable object, the frontend can invoke other variable object operations—for example to obtain or change the value of a variable object, or to change display format.
Variable objects have hierarchical tree structure. Any variable object that corresponds to a composite type, such as structure in C, has a number of child variable objects, for example corresponding to each element of a structure. A child variable object can itself have children, recursively. Recursion ends when we reach leaf variable objects, which always have built-in types. Child variable objects are created only by explicit request, so if a frontend is not interested in the children of a particular variable object, no child will be created.
For a leaf variable object it is possible to obtain its value as a string, or set the value from a string. String value can be also obtained for a non-leaf variable object, but it's generally a string that only indicates the type of the object, and does not list its contents. Assignment to a non-leaf variable object is not allowed.
A frontend does not need to read the values of all variable objects each time the program stops. Instead, MI provides an update command that lists all variable objects whose values has changed since the last update operation. This considerably reduces the amount of data that must be transferred to the frontend. As noted above, children variable objects are created on demand, and only leaf variable objects have a real value. As result, gdb will read target memory only for leaf variables that frontend has created.
The automatic update is not always desirable. For example, a frontend might want to keep a value of some expression for future reference, and never update it. For another example, fetching memory is relatively slow for embedded targets, so a frontend might want to disable automatic update for the variables that are either not visible on the screen, or “closed”. This is possible using so called “frozen variable objects”. Such variable objects are never implicitly updated.
Variable objects can be either fixed or floating. For the fixed variable object, the expression is parsed when the variable object is created, including associating identifiers to specific variables. The meaning of expression never changes. For a floating variable object the values of variables whose names appear in the expressions are re-evaluated every time in the context of the current frame. Consider this example:
void do_work(...) { struct work_state state; if (...) do_work(...); }
If a fixed variable object for the state
variable is created in
this function, and we enter the recursive call, the variable
object will report the value of state
in the top-level
do_work
invocation. On the other hand, a floating variable
object will report the value of state
in the current frame.
If an expression specified when creating a fixed variable object refers to a local variable, the variable object becomes bound to the thread and frame in which the variable object is created. When such variable object is updated, gdb makes sure that the thread/frame combination the variable object is bound to still exists, and re-evaluates the variable object in context of that thread/frame.
The following is the complete set of gdb/mi operations defined to access this functionality:
Operation | Description
|
-enable-pretty-printing
| enable Python-based pretty-printing
|
-var-create
| create a variable object
|
-var-delete
| delete the variable object and/or its children
|
-var-set-format
| set the display format of this variable
|
-var-show-format
| show the display format of this variable
|
-var-info-num-children
| tells how many children this object has
|
-var-list-children
| return a list of the object's children
|
-var-info-type
| show the type of this variable object
|
-var-info-expression
| print parent-relative expression that this variable object represents
|
-var-info-path-expression
| print full expression that this variable object represents
|
-var-show-attributes
| is this variable editable? does it exist here?
|
-var-evaluate-expression
| get the value of this variable
|
-var-assign
| set the value of this variable
|
-var-update
| update the variable and its children
|
-var-set-frozen
| set frozeness attribute
|
-var-set-update-range
| set range of children to display on update
|
In the next subsection we describe each operation in detail and suggest how it can be used.
-enable-pretty-printing
Command-enable-pretty-printing
gdb allows Python-based visualizers to affect the output of the MI variable object commands. However, because there was no way to implement this in a fully backward-compatible way, a front end must request that this functionality be enabled.
Once enabled, this feature cannot be disabled.
Note that if Python support has not been compiled into gdb, this command will still succeed (and do nothing).
This feature is currently (as of gdb 7.0) experimental, and may work differently in future versions of gdb.
-var-create
Command-var-create {name | "-"} {frame-addr | "*" | "@"} expression
This operation creates a variable object, which allows the monitoring of a variable, the result of an expression, a memory cell or a CPU register.
The name parameter is the string by which the object can be referenced. It must be unique. If `-' is specified, the varobj system will generate a string “varNNNNNN” automatically. It will be unique provided that one does not specify name of that format. The command fails if a duplicate name is found.
The frame under which the expression should be evaluated can be specified by frame-addr. A `*' indicates that the current frame should be used. A `@' indicates that a floating variable object must be created.
expression is any expression valid on the current language set (must not begin with a `*'), or one of the following:
A varobj's contents may be provided by a Python-based pretty-printer. In this
case the varobj is known as a dynamic varobj. Dynamic varobjs
have slightly different semantics in some cases. If the
-enable-pretty-printing
command is not sent, then gdb
will never create a dynamic varobj. This ensures backward
compatibility for existing clients.
This operation returns attributes of the newly-created varobj. These are:
struct
), or for a dynamic varobj, this value
will not be interesting.
on
, the
actual (derived) type of the object is shown rather than the
declared one.
display_hint
method. See Pretty Printing API.
Typical output will look like this:
name="name",numchild="N",type="type",thread-id="M", has_more="has_more"
-var-delete
Command-var-delete [ -c ] name
Deletes a previously created variable object and all of its children. With the `-c' option, just deletes the children.
Returns an error if the object name is not found.
-var-set-format
Command-var-set-format name format-spec
Sets the output format for the value of the object name to be format-spec.
The syntax for the format-spec is as follows:
format-spec ==> {binary | decimal | hexadecimal | octal | natural}
The natural format is the default format choosen automatically
based on the variable type (like decimal for an int
, hex
for pointers, etc.).
For a variable with children, the format is set only on the variable itself, and the children are not affected.
-var-show-format
Command-var-show-format name
Returns the format used to display the value of the object name.
format ==> format-spec
-var-info-num-children
Command-var-info-num-children name
Returns the number of children of a variable object name:
numchild=n
Note that this number is not completely reliable for a dynamic varobj. It will return the current number of children, but more children may be available.
-var-list-children
Command-var-list-children [print-values] name [from to]
Return a list of the children of the specified variable object and
create variable objects for them, if they do not already exist. With
a single argument or if print-values has a value of 0 or
--no-values
, print only the names of the variables; if
print-values is 1 or --all-values
, also print their
values; and if it is 2 or --simple-values
print the name and
value for simple data types and just the name for arrays, structures
and unions.
from and to, if specified, indicate the range of children to report. If from or to is less than zero, the range is reset and all children will be reported. Otherwise, children starting at from (zero-based) and up to and excluding to will be reported.
If a child range is requested, it will only affect the current call to
-var-list-children
, but not future calls to -var-update
.
For this, you must instead use -var-set-update-range
. The
intent of this approach is to enable a front end to implement any
update approach it likes; for example, scrolling a view may cause the
front end to request more children with -var-list-children
, and
then the front end could call -var-set-update-range
with a
different range to ensure that future updates are restricted to just
the visible items.
For each child the following results are returned:
For a dynamic varobj, this value cannot be used to form an expression. There is no way to do this at all with a dynamic varobj.
For C/C++ structures there are several pseudo children returned to designate access qualifiers. For these pseudo children exp is `public', `private', or `protected'. In this case the type and value are not present.
A dynamic varobj will not report the access qualifying
pseudo-children, regardless of the language. This information is not
available at all with a dynamic varobj.
on
, the
actual (derived) type of the object is shown rather than the
declared one.
The result may have its own attributes:
display_hint
method. See Pretty Printing API.
(gdb) -var-list-children n ^done,numchild=n,children=[child={name=name,exp=exp, numchild=n,type=type},(repeats N times)] (gdb) -var-list-children --all-values n ^done,numchild=n,children=[child={name=name,exp=exp, numchild=n,value=value,type=type},(repeats N times)]
-var-info-type
Command-var-info-type name
Returns the type of the specified variable name. The type is returned as a string in the same format as it is output by the gdb CLI:
type=typename
-var-info-expression
Command-var-info-expression name
Returns a string that is suitable for presenting this variable object in user interface. The string is generally not valid expression in the current language, and cannot be evaluated.
For example, if a
is an array, and variable object
A
was created for a
, then we'll get this output:
(gdb) -var-info-expression A.1 ^done,lang="C",exp="1"
Here, the values of lang
can be {"C" | "C++" | "Java"}
.
Note that the output of the -var-list-children
command also
includes those expressions, so the -var-info-expression
command
is of limited use.
-var-info-path-expression
Command-var-info-path-expression name
Returns an expression that can be evaluated in the current
context and will yield the same value that a variable object has.
Compare this with the -var-info-expression
command, which
result can be used only for UI presentation. Typical use of
the -var-info-path-expression
command is creating a
watchpoint from a variable object.
This command is currently not valid for children of a dynamic varobj, and will give an error when invoked on one.
For example, suppose C
is a C++ class, derived from class
Base
, and that the Base
class has a member called
m_size
. Assume a variable c
is has the type of
C
and a variable object C
was created for variable
c
. Then, we'll get this output:
(gdb) -var-info-path-expression C.Base.public.m_size ^done,path_expr=((Base)c).m_size)
-var-show-attributes
Command-var-show-attributes name
List attributes of the specified variable object name:
status=attr [ ( ,attr )* ]
where attr is { { editable | noneditable } | TBD }
.
-var-evaluate-expression
Command-var-evaluate-expression [-f format-spec] name
Evaluates the expression that is represented by the specified variable
object and returns its value as a string. The format of the string
can be specified with the `-f' option. The possible values of
this option are the same as for -var-set-format
(see -var-set-format). If the `-f' option is not specified,
the current display format will be used. The current display format
can be changed using the -var-set-format
command.
value=value
Note that one must invoke -var-list-children
for a variable
before the value of a child variable can be evaluated.
-var-assign
Command-var-assign name expression
Assigns the value of expression to the variable object specified
by name. The object must be `editable'. If the variable's
value is altered by the assign, the variable will show up in any
subsequent -var-update
list.
(gdb) -var-assign var1 3 ^done,value="3" (gdb) -var-update * ^done,changelist=[{name="var1",in_scope="true",type_changed="false"}] (gdb)
-var-update
Command-var-update [print-values] {name | "*"}
Reevaluate the expressions corresponding to the variable object
name and all its direct and indirect children, and return the
list of variable objects whose values have changed; name must
be a root variable object. Here, “changed” means that the result of
-var-evaluate-expression
before and after the
-var-update
is different. If `*' is used as the variable
object names, all existing variable objects are updated, except
for frozen ones (see -var-set-frozen). The option
print-values determines whether both names and values, or just
names are printed. The possible values of this option are the same
as for -var-list-children
(see -var-list-children). It is
recommended to use the `--all-values' option, to reduce the
number of MI commands needed on each program stop.
With the `*' parameter, if a variable object is bound to a currently running thread, it will not be updated, without any diagnostic.
If -var-set-update-range
was previously used on a varobj, then
only the selected range of children will be reported.
-var-update
reports all the changed varobjs in a tuple named
`changelist'.
Each item in the change list is itself a tuple holding:
"true"
"false"
"invalid"
file
command. The front end should normally choose to delete these variable
objects.
In the future new values may be added to this list so the front should
be prepared for this possibility. See GDB/MI Development and Front Ends.
When a varobj's type changes, its children are also likely to have
become incorrect. Therefore, the varobj's children are automatically
deleted when this attribute is `true'. Also, the varobj's update
range, when set using the -var-set-update-range
command, is
unset.
The `numchild' field in other varobj responses is generally not valid for a dynamic varobj – it will show the number of children that gdb knows about, but because dynamic varobjs lazily instantiate their children, this will not reflect the number of children which may be available.
The `new_num_children' attribute only reports changes to the
number of children known by gdb. This is the only way to
detect whether an update has removed children (which necessarily can
only happen at the end of the update range).
-var-set-update-range
), then they will
be listed in this attribute.
(gdb) -var-assign var1 3 ^done,value="3" (gdb) -var-update --all-values var1 ^done,changelist=[{name="var1",value="3",in_scope="true", type_changed="false"}] (gdb)
-var-set-frozen
Command-var-set-frozen name flag
Set the frozenness flag on the variable object name. The
flag parameter should be either `1' to make the variable
frozen or `0' to make it unfrozen. If a variable object is
frozen, then neither itself, nor any of its children, are
implicitly updated by -var-update
of
a parent variable or by -var-update *
. Only
-var-update
of the variable itself will update its value and
values of its children. After a variable object is unfrozen, it is
implicitly updated by all subsequent -var-update
operations.
Unfreezing a variable does not update it, only subsequent
-var-update
does.
(gdb) -var-set-frozen V 1 ^done (gdb)
-var-set-update-range
command-var-set-update-range name from to
Set the range of children to be returned by future invocations of
-var-update
.
from and to indicate the range of children to report. If from or to is less than zero, the range is reset and all children will be reported. Otherwise, children starting at from (zero-based) and up to and excluding to will be reported.
(gdb) -var-set-update-range V 1 2 ^done
-var-set-visualizer
command-var-set-visualizer name visualizer
Set a visualizer for the variable object name.
visualizer is the visualizer to use. The special value `None' means to disable any visualizer in use.
If not `None', visualizer must be a Python expression. This expression must evaluate to a callable object which accepts a single argument. gdb will call this object with the value of the varobj name as an argument (this is done so that the same Python pretty-printing code can be used for both the CLI and MI). When called, this object must return an object which conforms to the pretty-printing interface (see Pretty Printing API).
The pre-defined function gdb.default_visualizer
may be used to
select a visualizer by following the built-in process
(see Selecting Pretty-Printers). This is done automatically when
a varobj is created, and so ordinarily is not needed.
This feature is only available if Python support is enabled. The MI
command -list-features
(see GDB/MI Miscellaneous Commands)
can be used to check this.
Resetting the visualizer:
(gdb) -var-set-visualizer V None ^done
Reselecting the default (type-based) visualizer:
(gdb) -var-set-visualizer V gdb.default_visualizer ^done
Suppose SomeClass
is a visualizer class. A lambda expression
can be used to instantiate this class for a varobj:
(gdb) -var-set-visualizer V "lambda val: SomeClass()" ^done