Get desktop application:
View/edit binary Protocol Buffers messages
The Controller service provides the API for orchestrating a collection of debugger agents to perform debugging tasks. These agents are each attached to a process of an application which may include one or more replicas. The debugger agents register with the Controller to identify the application being debugged, the Debuggee. All agents that register with the same data, represent the same Debuggee, and are assigned the same `debuggee_id`. The debugger agents call the Controller to retrieve the list of active Breakpoints. Agents with the same `debuggee_id` get the same breakpoints list. An agent that can fulfill the breakpoint request updates the Controller with the breakpoint result. The controller selects the first result received and discards the rest of the results. Agents that poll again for active breakpoints will no longer have the completed breakpoint in the list and should remove that breakpoint from their attached process. The Controller service does not provide a way to retrieve the results of a completed breakpoint. This functionality is available using the Debugger service.
Registers the debuggee with the controller service. All agents attached to the same application must call this method with exactly the same request content to get back the same stable `debuggee_id`. Agents should call this method again whenever `google.rpc.Code.NOT_FOUND` is returned from any controller method. This protocol allows the controller service to disable debuggees, recover from data loss, or change the `debuggee_id` format. Agents must handle `debuggee_id` value changing upon re-registration.
Request to register a debuggee.
Required. Debuggee information to register. The fields `project`, `uniquifier`, `description` and `agent_version` of the debuggee must be set.
Response for registering a debuggee.
Debuggee resource. The field `id` is guaranteed to be set (in addition to the echoed fields). If the field `is_disabled` is set to `true`, the agent should disable itself by removing all breakpoints and detaching from the application. It should however continue to poll `RegisterDebuggee` until reenabled.
Returns the list of all active breakpoints for the debuggee. The breakpoint specification (`location`, `condition`, and `expressions` fields) is semantically immutable, although the field values may change. For example, an agent may update the location line number to reflect the actual line where the breakpoint was set, but this doesn't change the breakpoint semantics. This means that an agent does not need to check if a breakpoint has changed when it encounters the same breakpoint on a successive call. Moreover, an agent should remember the breakpoints that are completed until the controller removes them from the active list to avoid setting those breakpoints again.
Request to list active breakpoints.
Required. Identifies the debuggee.
A token that, if specified, blocks the method call until the list of active breakpoints has changed, or a server-selected timeout has expired. The value should be set from the `next_wait_token` field in the last response. The initial value should be set to `"init"`.
If set to `true` (recommended), returns `google.rpc.Code.OK` status and sets the `wait_expired` response field to `true` when the server-selected timeout has expired. If set to `false` (deprecated), returns `google.rpc.Code.ABORTED` status when the server-selected timeout has expired.
Response for listing active breakpoints.
List of all active breakpoints. The fields `id` and `location` are guaranteed to be set on each breakpoint.
A token that can be used in the next method call to block until the list of breakpoints changes.
If set to `true`, indicates that there is no change to the list of active breakpoints and the server-selected timeout has expired. The `breakpoints` field would be empty and should be ignored.
Updates the breakpoint state or mutable fields. The entire Breakpoint message must be sent back to the controller service. Updates to active breakpoint fields are only allowed if the new value does not change the breakpoint specification. Updates to the `location`, `condition` and `expressions` fields should not alter the breakpoint semantics. These may only make changes such as canonicalizing a value or snapping the location to the correct line of code.
Request to update an active breakpoint.
Required. Identifies the debuggee being debugged.
Required. Updated breakpoint information. The field `id` must be set. The agent must echo all Breakpoint specification fields in the update.
Response for updating an active breakpoint. The message is defined to allow future extensions.
(message has no fields)
The Debugger service provides the API that allows users to collect run-time information from a running application, without stopping or slowing it down and without modifying its state. An application may include one or more replicated processes performing the same work. A debugged application is represented using the Debuggee concept. The Debugger service provides a way to query for available debuggees, but does not provide a way to create one. A debuggee is created using the Controller service, usually by running a debugger agent with the application. The Debugger service enables the client to set one or more Breakpoints on a Debuggee and collect the results of the set Breakpoints.
Sets the breakpoint to the debuggee.
Request to set a breakpoint
Required. ID of the debuggee where the breakpoint is to be set.
Required. Breakpoint specification to set. The field `location` of the breakpoint must be set.
Required. The client version making the call. Schema: `domain/type/version` (e.g., `google.com/intellij/v1`).
Response for setting a breakpoint.
Breakpoint resource. The field `id` is guaranteed to be set (in addition to the echoed fileds).
Gets breakpoint information.
Request to get breakpoint information.
Required. ID of the debuggee whose breakpoint to get.
Required. ID of the breakpoint to get.
Required. The client version making the call. Schema: `domain/type/version` (e.g., `google.com/intellij/v1`).
Response for getting breakpoint information.
Complete breakpoint state. The fields `id` and `location` are guaranteed to be set.
Deletes the breakpoint from the debuggee.
Request to delete a breakpoint.
Required. ID of the debuggee whose breakpoint to delete.
Required. ID of the breakpoint to delete.
Required. The client version making the call. Schema: `domain/type/version` (e.g., `google.com/intellij/v1`).
Lists all breakpoints for the debuggee.
Request to list breakpoints.
Required. ID of the debuggee whose breakpoints to list.
When set to `true`, the response includes the list of breakpoints set by any user. Otherwise, it includes only breakpoints set by the caller.
When set to `true`, the response includes active and inactive breakpoints. Otherwise, it includes only active breakpoints.
When set, the response includes only breakpoints with the specified action.
This field is deprecated. The following fields are always stripped out of the result: `stack_frames`, `evaluated_expressions` and `variable_table`.
A wait token that, if specified, blocks the call until the breakpoints list has changed, or a server selected timeout has expired. The value should be set from the last response. The error code `google.rpc.Code.ABORTED` (RPC) is returned on wait timeout, which should be called again with the same `wait_token`.
Required. The client version making the call. Schema: `domain/type/version` (e.g., `google.com/intellij/v1`).
Response for listing breakpoints.
List of breakpoints matching the request. The fields `id` and `location` are guaranteed to be set on each breakpoint. The fields: `stack_frames`, `evaluated_expressions` and `variable_table` are cleared on each breakpoint regardless of its status.
A wait token that can be used in the next call to `list` (REST) or `ListBreakpoints` (RPC) to block until the list of breakpoints has changes.
Lists all the debuggees that the user has access to.
Request to list debuggees.
Required. Project number of a Google Cloud project whose debuggees to list.
When set to `true`, the result includes all debuggees. Otherwise, the result includes only debuggees that are active.
Required. The client version making the call. Schema: `domain/type/version` (e.g., `google.com/intellij/v1`).
Response for listing debuggees.
List of debuggees accessible to the calling user. The fields `debuggee.id` and `description` are guaranteed to be set. The `description` field is a human readable field provided by agents and can be displayed to users.
Represents the breakpoint specification, status and results.
Used in:
, , , , ,Breakpoint identifier, unique in the scope of the debuggee.
Action that the agent should perform when the code at the breakpoint location is hit.
Breakpoint source location.
Condition that triggers the breakpoint. The condition is a compound boolean expression composed using expressions in a programming language at the source location.
List of read-only expressions to evaluate at the breakpoint location. The expressions are composed using expressions in the programming language at the source location. If the breakpoint action is `LOG`, the evaluated expressions are included in log statements.
Only relevant when action is `LOG`. Defines the message to log when the breakpoint hits. The message may include parameter placeholders `$0`, `$1`, etc. These placeholders are replaced with the evaluated value of the appropriate expression. Expressions not referenced in `log_message_format` are not logged. Example: `Message received, id = $0, count = $1` with `expressions` = `[ message.id, message.count ]`.
Indicates the severity of the log. Only relevant when action is `LOG`.
When true, indicates that this is a final result and the breakpoint state will not change from here on.
Time this breakpoint was created by the server in seconds resolution.
Time this breakpoint was finalized as seen by the server in seconds resolution.
E-mail address of the user that created this breakpoint
Breakpoint status. The status includes an error flag and a human readable message. This field is usually unset. The message can be either informational or an error message. Regardless, clients should always display the text message back to the user. Error status indicates complete failure of the breakpoint. Example (non-final state): `Still loading symbols...` Examples (final state): * `Invalid line number` referring to location * `Field f not found in class C` referring to condition
The stack at breakpoint time, where stack_frames[0] represents the most recently entered function.
Values of evaluated expressions at breakpoint time. The evaluated expressions appear in exactly the same order they are listed in the `expressions` field. The `name` field holds the original expression text, the `value` or `members` field holds the result of the evaluated expression. If the expression cannot be evaluated, the `status` inside the `Variable` will indicate an error and contain the error text.
The `variable_table` exists to aid with computation, memory and network traffic optimization. It enables storing a variable once and reference it from multiple variables, including variables stored in the `variable_table` itself. For example, the same `this` object, which may appear at many levels of the stack, can have all of its data stored once in this table. The stack frame variables then would hold only a reference to it. The variable `var_table_index` field is an index into this repeated field. The stored objects are nameless and get their name from the referencing variable. The effective variable is a merge of the referencing variable and the referenced variable.
A set of custom breakpoint properties, populated by the agent, to be displayed to the user.
Actions that can be taken when a breakpoint hits. Agents should reject breakpoints with unsupported or unknown action values.
Used in:
,Capture stack frame and variables and update the breakpoint. The data is only captured once. After that the breakpoint is set in a final state.
Log each breakpoint hit. The breakpoint remains active until deleted or expired.
Log severity levels.
Used in:
Information log message.
Warning log message.
Error log message.
Represents the debugged application. The application may include one or more replicated processes executing the same code. Each of these processes is attached with a debugger agent, carrying out the debugging commands. Agents attached to the same debuggee identify themselves as such by using exactly the same Debuggee message value when registering.
Used in:
, ,Unique identifier for the debuggee generated by the controller service.
Project the debuggee is associated with. Use project number or id when registering a Google Cloud Platform project.
Uniquifier to further distinguish the application. It is possible that different applications might have identical values in the debuggee message, thus, incorrectly identified as a single application by the Controller service. This field adds salt to further distinguish the application. Agents should consider seeding this field with value that identifies the code, binary, configuration and environment.
Human readable description of the debuggee. Including a human-readable project name, environment name and version information is recommended.
If set to `true`, indicates that Controller service does not detect any activity from the debuggee agents and the application is possibly stopped.
Version ID of the agent. Schema: `domain/language-platform/vmajor.minor` (for example `google.com/java-gcp/v1.1`).
If set to `true`, indicates that the agent should disable itself and detach from the debuggee.
Human readable message to be displayed to the user about this debuggee. Absence of this field indicates no status. The message can be either informational or an error status.
References to the locations and revisions of the source code used in the deployed application.
References to the locations and revisions of the source code used in the deployed application.
A set of custom debuggee properties, populated by the agent, to be displayed to the user.
Represents a message with parameters.
Used in:
Format template for the message. The `format` uses placeholders `$0`, `$1`, etc. to reference parameters. `$$` can be used to denote the `$` character. Examples: * `Failed to load '$0' which helps debug $1 the first time it is loaded. Again, $0 is very important.` * `Please pay $$10 to use $0 instead of $1.`
Optional parameters to be embedded into the message.
Wrapper message for `Breakpoint.Action`. Defines a filter on the action field of breakpoints.
Used in:
Only breakpoints with the specified action will pass the filter.
Represents a location in the source code.
Used in:
,Path to the source file within the source context of the target binary.
Line inside the file. The first line in the file has the value `1`.
Column within a line. The first column in a line as the value `1`. Agents that do not support setting breakpoints on specific columns ignore this field.
Represents a stack frame context.
Used in:
Demangled function name at the call site.
Source location of the call site.
Set of arguments passed to this function. Note that this might not be populated for all stack frames.
Set of local variables at the stack frame location. Note that this might not be populated for all stack frames.
Represents a contextual status message. The message can indicate an error or informational status, and refer to specific parts of the containing object. For example, the `Breakpoint.status` field can indicate an error referring to the `BREAKPOINT_SOURCE_LOCATION` with the message `Location not found`.
Used in:
, ,Distinguishes errors from informational messages.
Reference to which the message applies.
Status message text.
Enumerates references to which the message applies.
Used in:
Status doesn't refer to any particular input.
Status applies to the breakpoint and is related to its location.
Status applies to the breakpoint and is related to its condition.
Status applies to the breakpoint and is related to its expressions.
Status applies to the breakpoint and is related to its age.
Status applies to the entire variable.
Status applies to variable value (variable name is valid).
Represents a variable or an argument possibly of a compound object type. Note how the following variables are represented: 1) A simple variable: int x = 5 { name: "x", value: "5", type: "int" } // Captured variable 2) A compound object: struct T { int m1; int m2; }; T x = { 3, 7 }; { // Captured variable name: "x", type: "T", members { name: "m1", value: "3", type: "int" }, members { name: "m2", value: "7", type: "int" } } 3) A pointer where the pointee was captured: T x = { 3, 7 }; T* p = &x; { // Captured variable name: "p", type: "T*", value: "0x00500500", members { name: "m1", value: "3", type: "int" }, members { name: "m2", value: "7", type: "int" } } 4) A pointer where the pointee was not captured: T* p = new T; { // Captured variable name: "p", type: "T*", value: "0x00400400" status { is_error: true, description { format: "unavailable" } } } The status should describe the reason for the missing value, such as `<optimized out>`, `<inaccessible>`, `<pointers limit reached>`. Note that a null pointer should not have members. 5) An unnamed value: int* p = new int(7); { // Captured variable name: "p", value: "0x00500500", type: "int*", members { value: "7", type: "int" } } 6) An unnamed pointer where the pointee was not captured: int* p = new int(7); int** pp = &p; { // Captured variable name: "pp", value: "0x00500500", type: "int**", members { value: "0x00400400", type: "int*" status { is_error: true, description: { format: "unavailable" } } } } } To optimize computation, memory and network traffic, variables that repeat in the output multiple times can be stored once in a shared variable table and be referenced using the `var_table_index` field. The variables stored in the shared table are nameless and are essentially a partition of the complete variable. To reconstruct the complete variable, merge the referencing variable with the referenced variable. When using the shared variable table, the following variables: T x = { 3, 7 }; T* p = &x; T& r = x; { name: "x", var_table_index: 3, type: "T" } // Captured variables { name: "p", value "0x00500500", type="T*", var_table_index: 3 } { name: "r", type="T&", var_table_index: 3 } { // Shared variable table entry #3: members { name: "m1", value: "3", type: "int" }, members { name: "m2", value: "7", type: "int" } } Note that the pointer address is stored with the referencing variable and not with the referenced variable. This allows the referenced variable to be shared between pointers and references. The type field is optional. The debugger agent may or may not support it.
Used in:
,Name of the variable, if any.
Simple value of the variable.
Variable type (e.g. `MyClass`). If the variable is split with `var_table_index`, `type` goes next to `value`. The interpretation of a type is agent specific. It is recommended to include the dynamic type rather than a static type of an object.
Members contained or pointed to by the variable.
Reference to a variable in the shared variable table. More than one variable can reference the same variable in the table. The `var_table_index` field is an index into `variable_table` in Breakpoint.
Status associated with the variable. This field will usually stay unset. A status of a single variable only applies to that variable or expression. The rest of breakpoint data still remains valid. Variables might be reported in error state even when breakpoint is not in final state. The message may refer to variable name with `refers_to` set to `VARIABLE_NAME`. Alternatively `refers_to` will be set to `VARIABLE_VALUE`. In either case variable value and members will be unset. Example of error message applied to name: `Invalid expression syntax`. Example of information message applied to value: `Not captured`. Examples of error message applied to value: * `Malformed string`, * `Field f not found in class C` * `Null pointer dereference`