5.6. How-To

5.6.1. Getting Started

The RTEMS specification items and qualification tools are work in progress. The first step to work with the RTEMS specification and the corresponding tools is a clone of the following repository:

git clone git://git.rtems.org/rtems-central.git
git submodule init
git submodule update

The tools need a virtual Python 3 environment. To set it up use:

cd rtems-central
make env

Each time you want to use one of the tools, you have to activate the environment in your shell:

cd rtems-central
. env/bin/activate

5.6.2. Application Configuration Options

The application configuration options and groups are maintained by specification items in the directory spec/if/acfg. Application configuration options are grouped by Application Configuration Group Item Type items which should be stored in files using the spec/if/acfg/group-*.yml pattern. Each application configuration option shall link to exactly one group item with the Application Configuration Group Member Link Role. There are four application option item types available which cover all existing options:

  • The feature enable options let the application enable a feature option. If the option is not defined, then the feature is simply not available or active. There should be no feature-specific code linked to the application if the option is not defined. Examples are options which enable a device driver like CONFIGURE_APPLICATION_NEEDS_CLOCK_DRIVER. These options are specified by Application Configuration Feature Enable Option Item Type items.

  • The feature options let the application enable a specific feature option. If the option is not defined, then a default feature option is used. Regardless whether the option is defined or not defined, feature-specific code may be linked to the application. Examples are options which disable a feature if the option is defined such as CONFIGURE_APPLICATION_DISABLE_FILESYSTEM and options which provide a stub implementation of a feature by default and a full implementation if the option is defined such as CONFIGURE_IMFS_ENABLE_MKFIFO. These options are specified by Application Configuration Feature Option Item Type items.

  • The integer value options let the application define a specific value for a system parameter. If the option is not defined, then a default value is used for the system parameter. Examples are options which define the maximum count of objects available for application use such as CONFIGURE_MAXIMUM_TASKS. These options are specified by Application Configuration Value Option Item Type items.

  • The initializer options let the application define a specific initializer for a system parameter. If the option is not defined, then a default setting is used for the system parameter. An example option of this type is CONFIGURE_INITIAL_EXTENSIONS. These options are specified by Application Configuration Value Option Item Type items.

Sphinx documentation sources and header files with Doxygen markup are generated from the specification items. The descriptions in the items shall use a restricted Sphinx formatting. Emphasis via one asterisk (“*”), strong emphasis via two asterisk (“**”), code samples via blockquotes (“``”), code blocks (“.. code-block:: c”) and lists are allowed. References to interface items are also allowed, for example “${appl-needs-clock-driver:/name}” and “${../rtems/tasks/create:/name}”. References to other parts of the documentation are possible, however, they are currently provided by hard-coded tables in rtemsspec/applconfig.py.

5.6.2.1. Modify an Existing Group

Search for the group by its section header and edit the specification item file. For example:

$ grep -rl "name: General System Configuration" spec/if/acfg
spec/if/acfg/group-general.yml
$ vi spec/if/acfg/group-general.yml

5.6.2.2. Modify an Existing Option

Search for the option by its C preprocessor define name and edit the specification item file. For example:

$ grep -rl CONFIGURE_APPLICATION_NEEDS_CLOCK_DRIVER spec/if/acfg
spec/if/acfg/appl-needs-clock-driver.yml
$ vi spec/if/acfg/appl-needs-clock-driver.yml

5.6.2.3. Add a New Group

Let new be the UID name part of the new group. Create the file spec/if/acfg/group-new.yml and provide all attributes for an Application Configuration Group Item Type item. For example:

$ vi spec/if/acfg/group-new.yml

5.6.2.4. Add a New Option

Let my-new-option be the UID name of the option. Create the file if/acfg/my-new-option.yml and provide all attributes for an appropriate refinement of Application Configuration Option Item Type. For example:

$ vi spec/if/acfg/my-new-option.yml

5.6.2.5. Generate Content after Changes

Once you are done with the modifications of an existing item or the creation of a new item, the changes need to be propagated to generated source files. This is done by the spec2modules.py script. Before you call this script, make sure the Git submodules are up-to-date.

$ ./spec2dmodules.py

The script modifies or creates source files in modules/rtems and modules/rtems-docs. Create patch sets for these changes just as if these were work done by a human and follow the normal patch review process described in the RTEMS User Manual. When the changes are integrated, update the Git submodules and check in the changed items.

5.6.3. Glossary Specification

The glossary of terms for the RTEMS Project is defined by Glossary Term Item Type items in the spec/glossary directory. For a new glossary term add a glossary item to this directory. As the file name use the term in lower case with all white space and special characters removed or replaced by alphanumeric characters, for example spec/glossary/magicpower.yml for the term magic power.

Use ${uid:/attribute} substitutions to reference other parts of the specification.

SPDX-License-Identifier: CC-BY-SA-4.0 OR BSD-2-Clause
copyrights:
- Copyright (C) 2020 embedded brains GmbH (http://www.embedded-brains.de)
enabled-by: true
glossary-type: term
links:
- role: glossary-member
  uid: ../glossary-general
term: magic power
text: |
  Magic power enables a caller to create magic objects using a
  ${magicwand:/term}.
type: glossary

Define acronyms with the phrase This term is an acronym for *. in the text attribute:

...
term: MP
...
text: |
  This term is an acronym for Magic Power.
...

Once you are done with the glossary items, run the script spec2modules.py to generate the derived documentation content. Send patches for the generated documentation and the specification to the Developers Mailing List and follow the normal patch review process.

5.6.4. Interface Specification

5.6.4.1. Specify an API Header File

The RTEMS API header files are specified under spec:/if/rtems/*. Create a subdirectory with a corresponding name for the API, for example in spec/if/rtems/foo for the foo API. In this new subdirectory place an Interface Header File Item Type item named header.yml (spec/if/rtems/foo/header.yml) and populate it with the required attributes.

SPDX-License-Identifier: CC-BY-SA-4.0 OR BSD-2-Clause
copyrights:
- Copyright (C) 2020 embedded brains GmbH (http://www.embedded-brains.de)
enabled-by: true
interface-type: header-file
links:
- role: interface-placement
  uid: /if/domains/api
path: rtems/rtems/foo.h
prefix: cpukit/include
type: interface

5.6.4.2. Specify an API Element

Figure out the corresponding header file item. If it does not exist, see Specify an API Header File. Place a specialization of an Interface Item Type item into the directory of the header file item, for example spec/if/rtems/foo/bar.yml for the bar() function. Add the required attributes for the new interface item. Do not hard code interface names which are used to define the new interface. Use ${uid-of-interface-item:/name} instead. If the referenced interface is specified in the same directory, then use a relative UID. Using interface references creates implicit dependencies and helps the header file generator to resolve the interface dependencies and header file includes for you. Use Interface Unspecified Item Type items for interface dependencies to other domains such as the C language, the compiler, the implementation, or user-provided defines. To avoid cyclic dependencies between types you may use an Interface Forward Declaration Item Type item.

SPDX-License-Identifier: CC-BY-SA-4.0 OR BSD-2-Clause
brief: Tries to create a magic object and returns it.
copyrights:
- Copyright (C) 2020 embedded brains GmbH (http://www.embedded-brains.de)
definition:
  default:
    body: null
    params:
    - ${magic-wand:/name} ${.:/params[0]/name}
    return: ${magic-type:/name} *
  variants: []
description: |
  The magic object is created out of nothing with the help of a magic wand.
enabled-by: true
interface-type: function
links:
- role: interface-placement
  uid: header
- role: interface-ingroup
  uid: /groups/api/classic/foo
name: bar
notes: null
params:
- description: is the magic wand.
  dir: null
  name: magic_wand
return:
  return: Otherwise, the magic object is returned.
  return-values:
  - description: The caller did not have enough magic power.
    value: ${/if/c/null}
type: interface

5.6.5. Requirements Depending on Build Configuration Options

Use the enabled-by attribute of items or parts of an item to make it dependent on build configuration options such as RTEMS_SMP or architecture-specific options such as CPU_ENABLE_ROBUST_THREAD_DISPATCH, see Enabled-By Expression. With this attribute the specification can be customized at the level of an item or parts of an item. If the enabled-by attribute evaluates to false for a particular configuration, then the item or the associated part is disabled in the specification. The enabled-by attribute acts as a formalized where clause, see recommended requirements syntax.

Please have a look at the following example which specifies the transition map of rtems_signal_catch():

transition-map:
- enabled-by: true
  post-conditions:
    Status: Ok
    ASRInfo:
    - if:
        pre-conditions:
          Handler: Valid
      then: New
    - else: Inactive
  pre-conditions:
    Pending: all
    Handler: all
    Preempt: all
    Timeslice: all
    ASR: all
    IntLvl: all
- enabled-by: CPU_ENABLE_ROBUST_THREAD_DISPATCH
  post-conditions:
    Status: NotImplIntLvl
    ASRInfo: NopIntLvl
  pre-conditions:
    Pending: all
    Handler:
    - Valid
    Preempt: all
    Timeslice: all
    ASR: all
    IntLvl:
    - Positive
- enabled-by: RTEMS_SMP
  post-conditions:
    Status: NotImplNoPreempt
    ASRInfo: NopNoPreempt
  pre-conditions:
    Pending: all
    Handler:
    - Valid
    Preempt:
    - 'No'
    Timeslice: all
    ASR: all
    IntLvl: all

5.6.6. Requirements Depending on Application Configuration Options

Requirements which depend on application configuration options such as CONFIGURE_MAXIMUM_PROCESSORS should be written in the following syntax:

Where <feature is included>, the <system name> shall <system response>.

Use these clauses with care. Make sure all feature combinations are covered. Using a truth table may help. If a requirement depends on multiple features, use:

Where <feature 0>, where <feature 1>, where <feature …>, the <system name> shall <system response>.

For application configuration options, use the clauses like this:

CONFIGURE_MAXIMUM_PROCESSORS equal to one

Where the system was configured with a processor maximum of exactly one, …

CONFIGURE_MAXIMUM_PROCESSORS greater than one

Where the system was configured with a processor maximum greater than one, …

Please have a look at the following example used to specify rtems_signal_catch(). The example is a post-condition state specification of an action requirement, so there is an implicit set of pre-conditions and the trigger:

While <pre-condition(s)>, when rtems_signal_catch() is called, …

The where clauses should be mentally placed before the while clauses.

post-conditions:
- name: ASRInfo
  states:
  - name: NopNoPreempt
    test-code: |
      if ( rtems_configuration_get_maximum_processors() > 1 ) {
        CheckNoASRChange( ctx );
      } else {
        CheckNewASRSettings( ctx );
      }
    text: |
      Where the scheduler does not support the no-preempt mode, the ASR
      information of the caller of ${../if/catch:/name} shall not be
      changed by the ${../if/catch:/name} call.

      Where the scheduler does support the no-preempt mode, the ASR
      processing for the caller of ${../if/catch:/name} shall be done using
      the handler specified by ${../if/catch:/params[0]/name} in the mode
      specified by ${../if/catch:/params[1]/name}.

5.6.7. Action Requirements

Action Requirement Item Type items may be used to specify and validate directive calls. They are a generator for event-driven requirements. Event-driven requirements should be written in the following syntax:

While <pre-condition 0>, while <pre-condition 1>, …, while <pre-condition n>, when <trigger>, the <system name> shall <system response>.

The list of while <pre-condition i> clauses for i from 1 to n in the EARS notation is generated by n pre-condition states in the action requirement item, see the pre-condition attribute in the Action Requirement Item Type.

The <trigger> in the EARS notation is defined for an action requirement item by the link to an Interface Function Item Type or an Interface Macro Item Type item using the Interface Function Link Role. The code provided by the test-action attribute defines the action code which should invoke the trigger directive in a particular set of pre-condition states.

Each post-condition state of the action requirement item generates a <system name> shall <system response> clause in the EARS notation, see the post-condition attribute in the Action Requirement Item Type.

Each entry in the transition map is an event-driven requirement composed of the pre-condition states, the trigger defined by the link to a directive, and the post-condition states. The transition map is defined by a list of Action Requirement Transition descriptors.

Use CamelCase for the pre-condition names, post-condition names, and state names in action requirement items. The more conditions a directive has, the shorter should be the names. The transition map may be documented as a table and more conditions need more table columns. Use item attribute references in the text attributes. This allows context-sensitive substitutions.

5.6.7.1. Example

Lets have a look at an example of an action requirement item. We would like to specify and validate the behaviour of the

rtems_status_code rtems_timer_create( rtems_name name, rtems_id *id );

directive which is particularly simple. For a more complex example see the specification of rtems_signal_catch() or rtems_signal_send() in spec:/rtems/signal/req/catch or spec:/rtems/signal/send respectively.

The event triggers are calls to rtems_timer_create(). Firstly, we need the list of pre-conditions relevant to this directive. Good candidates are the directive parameters, this gives us the Name and Id conditions. A system condition is if an inactive timer object is available so that we can create a timer, this gives us the Free condition. Secondly, we need the list of post-conditions relevant to this directive. They are the return status of the directive, Status, the validity of a unique object name, Name, and the value of an object identifier variable, IdVar. Each condition has a set of states, see the YAML data below for the details. The specified conditions and states yield the following transition map:

Entry

Descriptor

Name

Id

Free

Status

Name

IdVar

0

0

Valid

Valid

Yes

Ok

Valid

Set

1

0

Valid

Valid

No

TooMany

Invalid

Nop

2

0

Valid

Null

Yes

InvAddr

Invalid

Nop

3

0

Valid

Null

No

InvAddr

Invalid

Nop

4

0

Invalid

Valid

Yes

InvName

Invalid

Nop

5

0

Invalid

Valid

No

InvName

Invalid

Nop

6

0

Invalid

Null

Yes

InvName

Invalid

Nop

7

0

Invalid

Null

No

InvName

Invalid

Nop

Not all transition maps are that small, the transition map of rtems_task_mode() has more than 8000 entries. We can construct requirements from the clauses of the entries. For example, the three requirements of entry 0 (Name=Valid, Id=Valid, and Free=Yes results in Status=Ok, Name=Valid, and IdVar=Set) are:

While the name parameter is valid, while the id parameter references an object of type rtems_id, while the system has at least one inactive timer object available, when rtems_timer_create() is called, the return status of rtems_timer_create() shall be RTEMS_SUCCESSFUL.

While the name parameter is valid, while the id parameter references an object of type rtems_id, while the system has at least one inactive timer object available, when rtems_timer_create() is called, the unique object name shall identify the timer created by the rtems_timer_create() call.

While the name parameter is valid, while the id parameter references an object of type rtems_id, while the system has at least one inactive timer object available, when rtems_timer_create() is called, the value of the object referenced by the id parameter shall be set to the object identifier of the created timer after the return of the rtems_timer_create() call.

Now we will have a look at the specification item line by line. The top-level attributes are normally in alphabetical order in an item file. For this presentation we use a structured order.

SPDX-License-Identifier: CC-BY-SA-4.0 OR BSD-2-Clause
copyrights:
- Copyright (C) 2021 embedded brains GmbH (http://www.embedded-brains.de)
enabled-by: true
functional-type: action
rationale: null
references: []
requirement-type: functional

The specification items need a bit of boilerplate to tell you what they are, who wrote them, and what their license is.

text: ${.:text-template}

Each requirement item needs a text attribute. For the action requirements, we do not have a single requirement. There is just a template indicator and no plain text. Several event-driven requirements are defined by the pre-conditions, the trigger, and the post-conditions.

pre-conditions:
- name: Name
  states:
  - name: Valid
    test-code: |
      ctx->name = NAME;
    text: |
      While the ${../if/create:/params[0]/name} parameter is valid.
  - name: Invalid
    test-code: |
      ctx->name = 0;
    text: |
      While the ${../if/create:/params[0]/name} parameter is invalid.
  test-epilogue: null
  test-prologue: null
- name: Id
  states:
  - name: Valid
    test-code: |
      ctx->id = &ctx->id_value;
    text: |
      While the ${../if/create:/params[1]/name} parameter references an object
      of type ${../../type/if/id:/name}.
  - name: 'Null'
    test-code: |
      ctx->id = NULL;
    text: |
      While the ${../if/create:/params[1]/name} parameter is
      ${/c/if/null:/name}.
  test-epilogue: null
  test-prologue: null
- name: Free
  states:
  - name: 'Yes'
    test-code: |
      /* Ensured by the test suite configuration */
    text: |
      While the system has at least one inactive timer object available.
  - name: 'No'
    test-code: |
      ctx->seized_objects = T_seize_objects( Create, NULL );
    text: |
      While the system has no inactive timer object available.
  test-epilogue: null
  test-prologue: null

This list defines the pre-conditions. Each pre-condition has a list of states and corresponding validation test code.

links:
- role: interface-function
  uid: ../if/create
test-action: |
  ctx->status = rtems_timer_create( ctx->name, ctx->id );

The link to the rtems_timer_create() interface specification item with the interface-function link role defines the trigger. The test-action defines the how the triggering directive is invoked for the validation test depending on the pre-condition states. The code is not always as simple as in this example. The validation test is defined in this item along with the specification.

post-conditions:
- name: Status
  states:
  - name: Ok
    test-code: |
      T_rsc_success( ctx->status );
    text: |
      The return status of ${../if/create:/name} shall be
      ${../../status/if/successful:/name}.
  - name: InvName
    test-code: |
      T_rsc( ctx->status, RTEMS_INVALID_NAME );
    text: |
      The return status of ${../if/create:/name} shall be
      ${../../status/if/invalid-name:/name}.
  - name: InvAddr
    test-code: |
      T_rsc( ctx->status, RTEMS_INVALID_ADDRESS );
    text: |
      The return status of ${../if/create:/name} shall be
      ${../../status/if/invalid-address:/name}.
  - name: TooMany
    test-code: |
      T_rsc( ctx->status, RTEMS_TOO_MANY );
    text: |
      The return status of ${../if/create:/name} shall be
      ${../../status/if/too-many:/name}.
  test-epilogue: null
  test-prologue: null
- name: Name
  states:
  - name: Valid
    test-code: |
      id = 0;
      sc = rtems_timer_ident( NAME, &id );
      T_rsc_success( sc );
      T_eq_u32( id, ctx->id_value );
    text: |
      The unique object name shall identify the timer created by the
      ${../if/create:/name} call.
  - name: Invalid
    test-code: |
      sc = rtems_timer_ident( NAME, &id );
      T_rsc( sc, RTEMS_INVALID_NAME );
    text: |
      The unique object name shall not identify a timer.
  test-epilogue: null
  test-prologue: |
    rtems_status_code sc;
    rtems_id          id;
- name: IdVar
  states:
  - name: Set
    test-code: |
      T_eq_ptr( ctx->id, &ctx->id_value );
      T_ne_u32( ctx->id_value, INVALID_ID );
    text: |
      The value of the object referenced by the ${../if/create:/params[1]/name}
      parameter shall be set to the object identifier of the created timer
      after the return of the ${../if/create:/name} call.
  - name: Nop
    test-code: |
      T_eq_u32( ctx->id_value, INVALID_ID );
    text: |
      Objects referenced by the ${../if/create:/params[1]/name} parameter in
      past calls to ${../if/create:/name} shall not be accessed by the
      ${../if/create:/name} call.
  test-epilogue: null
  test-prologue: null

This list defines the post-conditions. Each post-condition has a list of states and corresponding validation test code.

skip-reasons: {}
transition-map:
- enabled-by: true
  post-conditions:
    Status:
    - if:
        pre-conditions:
          Name: Invalid
      then: InvName
    - if:
        pre-conditions:
          Id: 'Null'
      then: InvAddr
    - if:
        pre-conditions:
          Free: 'No'
      then: TooMany
    - else: Ok
    Name:
    - if:
        post-conditions:
          Status: Ok
      then: Valid
    - else: Invalid
    IdVar:
    - if:
        post-conditions:
          Status: Ok
      then: Set
    - else: Nop
  pre-conditions:
    Name: all
    Id: all
    Free: all
type: requirement

This list of transition descriptors defines the transition map. For the post-conditions, you can use expressions to ease the specification, see Action Requirement Transition Post-Condition State. The skip-reasons can be used to skip entire entries in the transition map, see Action Requirement Skip Reasons.

test-brief: null
test-description: null

The item contains the validation test code. The validation test in general can be described by these two attributes.

test-target: testsuites/validation/tc-timer-create.c

This is the target file for the generated validation test code. Make sure this file is included in the build specification, otherwise the test code generation will fail.

test-includes:
- rtems.h
- string.h
test-local-includes: []

You can specify a list of includes for the validation test.

test-header: null

A test header may be used to create a parameterized validation test, see Test Header. This is an advanced topic, see the specification of rtems_task_ident() for an example.

test-context-support: null
test-context:
- brief: |
    This member is used by the T_seize_objects() and T_surrender_objects()
    support functions.
  description: null
  member: |
    void *seized_objects
- brief: |
    This member may contain the object identifier returned by
    rtems_timer_create().
  description: null
  member: |
    rtems_id id_value
- brief: |
    This member specifies the ${../if/create:/params[0]/name} parameter for the
    action.
  description: null
  member: |
    rtems_name name
- brief: |
    This member specifies the ${../if/create:/params[1]/name} parameter for the
    action.
  description: null
  member: |
    rtems_id *id
- brief: |
    This member contains the return status of the action.
  description: null
  member: |
    rtems_status_code status

You can specify a list of validation test context members which can be used to maintain the state of the validation test. The context is available through an implicit ctx variable in all code blocks except the support blocks. The context support code can be used to define test-specific types used by context members. Do not use global variables.

test-support: |
  #define NAME rtems_build_name( 'T', 'E', 'S', 'T' )

  #define INVALID_ID 0xffffffff

  static rtems_status_code Create( void *arg, uint32_t *id )
  {
    return rtems_timer_create( rtems_build_name( 'S', 'I', 'Z', 'E' ), id );
  }

The support code block can be used to provide functions, data structures, and constants for the validation test.

test-prepare: null
test-cleanup: |
  if ( ctx->id_value != INVALID_ID ) {
    rtems_status_code sc;

    sc = rtems_timer_delete( ctx->id_value );
    T_rsc_success( sc );

    ctx->id_value = INVALID_ID;
  }

  T_surrender_objects( &ctx->seized_objects, rtems_timer_delete );

The validation test basically executes a couple of nested for loops to iterate over each pre-condition and each state of the pre-conditions. These two optional code blocks can be used to prepare the pre-condition state preparations and clean up after the post-condition checks in each loop iteration.

test-setup:
  brief: null
  code: |
    memset( ctx, 0, sizeof( *ctx ) );
    ctx->id_value = INVALID_ID;
  description: null
test-stop: null
test-teardown: null

These optional code blocks correspond to test fixture methods, see Test Fixture.

5.6.7.2. Pre-Condition Templates

Specify all directive parameters as separate pre-conditions. Use the following syntax for directive object identifier parameters:

- name: Id
  states:
  - name: NoObj
    test-code: |
      ctx->id = 0xffffffff;
    text: |
      While the ${../if/directive:/params[0]/name} parameter is not
      associated with a thing.
  - name: ClassA
    test-code: |
      ctx->id = ctx->class_a_id;
    text: |
      While the ${../if/directive:/params[0]/name} parameter is associated
      with a class A thing.
  - name: ClassB
    test-code: |
      ctx->id = ctx->class_b_id;
    text: |
      While the ${../if/directive:/params[0]/name} parameter is associated
      with a class B thing.
  test-epilogue: null
  test-prologue: null

Do not add specifications for invalid pointers. In general, there are a lot of invalid pointers and the use of an invalid pointer is in almost all cases undefined behaviour in RTEMS. There may be specifications for special cases which deal with some very specific invalid pointers such as the NULL pointer or pointers which do not satisfy a range or boundary condition. Use the following syntax for directive pointer parameters:

- name: Id
  states:
  - name: Valid
    test-code: |
      ctx->id = &ctx->id_value;
    text: |
      While the ${../if/directive:/params[3]/name} parameter references an
      object of type ${../../type/if/id:/name}.
  - name: 'Null'
    test-code: |
      ctx->id = NULL;
    text: |
      While the ${../if/directive:/params[3]/name} parameter is
      ${/c/if/null:/name}.
  test-epilogue: null
  test-prologue: null

Use the following syntax for other directive parameters:

- name: Name
  states:
  - name: Valid
    test-code: |
      ctx->name = NAME;
    text: |
      While the ${../if/directive:/params[0]/name} parameter is valid.
  - name: Invalid
    test-code: |
      ctx->name = 0;
    text: |
      While the ${../if/directive:/params[0]/name} parameter is invalid.
  test-epilogue: null
  test-prologue: null

5.6.7.3. Post-Condition Templates

Do not mix different things into one post-condition. If you write multiple sentences to describe what happened, then think about splitting up the post-condition. Keep the post-condition simple and focus on one testable aspect which may be changed by a directive call.

For directives returning an rtems_status_code use the following post-condition states. Specify only status codes which may be returned by the directive. Use it as the first post-condition. The first state shall be Ok. The other states shall be listed in the order in which they can occur.

- name: Status
  states:
  - name: Ok
    test-code: |
      T_rsc_success( ctx->status );
    text: |
      The return status of ${../if/directive:/name} shall be
      ${../../status/if/successful:/name}.
  - name: IncStat
    test-code: |
      T_rsc( ctx->status, RTEMS_INCORRECT_STATE );
    text: |
      The return status of ${../if/directive:/name} shall be
      ${../../status/if/incorrect-state:/name}.
  - name: InvAddr
    test-code: |
      T_rsc( ctx->status, RTEMS_INVALID_ADDRESS );
    text: |
      The return status of ${../if/directive:/name} shall be
      ${../../status/if/invalid-address:/name}.
  - name: InvName
    test-code: |
      T_rsc( ctx->status, RTEMS_INVALID_NAME );
    text: |
      The return status of ${../if/directive:/name} shall be
      ${../../status/if/invalid-name:/name}.
  - name: InvNum
    test-code: |
      T_rsc( ctx->status, RTEMS_INVALID_NUMBER );
    text: |
      The return status of ${../if/directive:/name} shall be
      ${../../status/if/invalid-number:/name}.
  - name: InvSize
    test-code: |
      T_rsc( ctx->status, RTEMS_INVALID_SIZE );
    text: |
      The return status of ${../if/directive:/name} shall be
      ${../../status/if/invalid-size:/name}.
  - name: InvPrio
    test-code: |
      T_rsc( ctx->status, RTEMS_INVALID_PRIORITY );
    text: |
      The return status of ${../if/directive:/name} shall be
      ${../../status/if/invalid-priority:/name}.
  - name: NotConf
    test-code: |
      T_rsc( ctx->status, RTEMS_NOT_CONFIGURED );
    text: |
      The return status of ${../if/directive:/name} shall be
      ${../../status/if/not-configured:/name}.
  - name: NotDef
    test-code: |
      T_rsc( ctx->status, RTEMS_NOT_DEFINED );
    text: |
      The return status of ${../if/directive:/name} shall be
      ${../../status/if/not-defined:/name}.
  - name: NotImpl
    test-code: |
      T_rsc( ctx->status, RTEMS_NOT_IMPLEMENTED );
    text: |
      The return status of ${../if/directive:/name} shall be
      ${../../status/if/not-implemented:/name}.
  - name: TooMany
    test-code: |
      T_rsc( ctx->status, RTEMS_TOO_MANY );
    text: |
      The return status of ${../if/directive:/name} shall be
      ${../../status/if/too-many:/name}.
  - name: Unsat
    test-code: |
      T_rsc( ctx->status, RTEMS_UNSATISFIED  );
    text: |
      The return status of ${../if/directive:/name} shall be
      ${../../status/if/unsatisfied:/name}.
  test-epilogue: null
  test-prologue: null

For values which are returned by reference through directive parameters, use the following post-condition states.

- name: SomeParamVar
  states:
  - name: Set
    test-code: |
      /* Add code to check that the object value was set to X */
    text: |
      The value of the object referenced by the
      ${../if/directive:/params[0]/name} parameter shall be set to X after
      the return of the ${../if/directive:/name} call.
  - name: Nop
    test-code: |
      /* Add code to check that the object was not modified */
    text: |
      Objects referenced by the ${../if/directive:/params[0]/name}
      parameter in past calls to ${../if/directive:/name} shall not be
      accessed by the ${../if/directive:/name} call.