Selective modelling of volatile variables #5374

danpoe · 2020-06-04T17:05:52Z

Each commit message has a non-empty body, explaining why the change was made.
Methods or procedures I have added are documented, following the guidelines provided in CODING_STANDARD.md.
The feature or user visible behaviour I have added or modified has been documented in the User Guide in doc/cprover-manual/
Regression or unit tests are included, or existing tests cover the modified code (in this case I have detailed which ones those are in the commit message).
n/a My commit message includes data points confirming performance improvements (if claimed).
My PR is restricted to a single feature or bugfix.
n/a White-space or formatting changes outside the feature-related changed lines are in commits of their own.

hannes-steffenhagen-diffblue

I'm OK with the code changes, if you have the time I think the commits themselves could be cleaned up a bit.

hannes-steffenhagen-diffblue · 2020-06-08T08:50:09Z

regression/goto-instrument/nondet-volatile-model-01/test.c

+#include <assert.h>
+
+volatile int a;
+volatile int b;


I think one of these should have a regression test involving the interaction with non-volatile variables; E.g.

volatile int x; int y = x; assert(x == y); // fails because volatile

hannes-steffenhagen-diffblue · 2020-06-08T09:34:36Z

src/goto-instrument/nondet_volatile.cpp

@@ -20,7 +20,29 @@ Date: September 2011

 using havoc_predicatet = std::function<bool(const exprt &)>;

-static bool is_volatile(const namespacet &ns, const typet &src)
+class nondet_volatilet


So... this is just a namespace? Maybe the commit should say why you’re doing this.

Mentioned in the commit message that this is in preparation for changes to come.

hannes-steffenhagen-diffblue · 2020-06-08T09:36:42Z

src/goto-instrument/nondet_volatile.cpp

+  if(ns.lookup(id, symbol))
+  {
+    throw invalid_command_line_argument_exceptiont(
+      "given name `" + id2string(id) + "` not found in symbol table",


⛏️ Should this also be "symbol" rather than "given name"?

hannes-steffenhagen-diffblue · 2020-06-08T09:38:20Z

src/goto-instrument/nondet_volatile.cpp

+  {
+    throw invalid_command_line_argument_exceptiont(
+      "given name `" + id2string(id) + "` not found in symbol table",
+      NONDET_VOLATILE_VARIABLE_OPT);


⛏️ Usually we start these with -- because the _OPT things don’t have it; Without the -- it might be slightly harder to recognise that this is referring to a command line option.

hannes-steffenhagen-diffblue · 2020-06-08T10:01:31Z

src/goto-instrument/nondet_volatile.cpp

+      NONDET_VOLATILE_VARIABLE_OPT);
+  }
+
+  if(!symbol->is_static_lifetime || !symbol->type.get_bool(ID_C_volatile))


It's a bit unclear to me why this only works with global variables.

Since the volatile variables will typically model hardware resources, it would probably make less sense to have volatile variables with automatic storage duration.

hannes-steffenhagen-diffblue · 2020-06-08T13:29:45Z

regression/goto-instrument/nondet-volatile-model-03/test.c

+
+void main()
+{
+  int result = a + b;


Also, does this work if the global is read via volatile pointer? (I'm assuming no, just for clarification).

Yes that's right, it currently only works for direct variable reads.

(One could also mark a pointer as nondet, in which case the pointer dereferences would be nondet as well. Though when --pointer-check would be enabled it would fail.)

hannes-steffenhagen-diffblue · 2020-06-08T14:07:57Z

src/goto-instrument/nondet_volatile.cpp

  const optionst &options)
 {
+  PRECONDITION(!all_nondet);


⛏️ There's a lot happening in 782395c which makes it a bit hard to recognise as a refactoring (i.e. there's so much code moving around its a bit hard to see where what is moving).

hannes-steffenhagen-diffblue · 2020-06-08T14:13:01Z

src/goto-instrument/nondet_volatile.cpp

@@ -101,27 +108,67 @@ bool nondet_volatilet::is_volatile(const namespacet &ns, const typet &src)

 void nondet_volatilet::handle_volatile_expression(
  exprt &expr,
-  const namespacet &ns)
+  const namespacet &ns,


⛏️ bd81144

Says it adds a new --nondet-volatile-model option, but this option is actually added in a different commit.

Changed commit message.

hannes-steffenhagen-diffblue · 2020-06-08T14:15:41Z

src/goto-instrument/nondet_volatile.cpp

 {
  if(expr.id()==ID_if)
  {
-    nondet_volatile_rhs(symbol_table, to_if_expr(expr).cond(), should_havoc);
+    nondet_volatile_rhs(symbol_table, to_if_expr(expr).cond());


❓ Is this case match already covered by existing tests somewhere?

This is currently untested as the original nondet volatile code on which this builds was untested. I'll add more tests for the existing code in a separate PR.

This is in preparation of adding a new feature to selectively provide models for volatile reads. For this member variables will be added to nondet_volatilet to store configuration.

This refactors the command line handling and typechecking in class nondet_volatilet, and adds a new option --nondet-volatile-model (unimplemented) which will allow selectivly providing models for volatile reads.

Refactor the code such as to allow later addition of new feature to selectively provide models for volatile reads

Implements the option --nondet-volatile-model <variable>:<model> to allow to specify models which are called whenever the given volatile variable is read. The return value of the model is used as the value read from the volatile variable.

codecov · 2020-06-09T12:38:00Z

Codecov Report

Merging #5374 into develop will increase coverage by 0.01%.
The diff coverage is 85.95%.

@@             Coverage Diff             @@
##           develop    #5374      +/-   ##
===========================================
+ Coverage    68.16%   68.17%   +0.01%     
===========================================
  Files         1173     1173              
  Lines        97401    97461      +60     
===========================================
+ Hits         66390    66445      +55     
- Misses       31011    31016       +5

Flag	Coverage Δ
#cproversmt2	`42.47% <ø> (ø)`
#regression	`65.38% <85.95%> (+0.01%)`	⬆️
#unit	`32.18% <ø> (ø)`

Impacted Files	Coverage Δ
src/goto-instrument/nondet_volatile.cpp	`86.80% <85.95%> (+3.47%)`	⬆️

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 8b604a7...d8c6441. Read the comment docs.

danpoe force-pushed the feature/selective-modelling-of-volatile-variables branch 3 times, most recently from 66d99c4 to d13a50b Compare June 8, 2020 08:25

danpoe marked this pull request as ready for review June 8, 2020 08:26

danpoe requested review from chrisr-diffblue, hannes-steffenhagen-diffblue, martin-cs, peterschrammel and smowton as code owners June 8, 2020 08:26

danpoe added the aws Bugs or features of importance to AWS CBMC users label Jun 8, 2020

hannes-steffenhagen-diffblue approved these changes Jun 8, 2020

View reviewed changes

danpoe added 7 commits June 9, 2020 11:42

Move nondet volatile functionality into class nondet_volatilet

611f8a1

This is in preparation of adding a new feature to selectively provide models for volatile reads. For this member variables will be added to nondet_volatilet to store configuration.

Rework command line handling of the nonde volatile feature

4b841c7

This refactors the command line handling and typechecking in class nondet_volatilet, and adds a new option --nondet-volatile-model (unimplemented) which will allow selectivly providing models for volatile reads.

Refactor existing nondet volatile code

47bb11e

Refactor the code such as to allow later addition of new feature to selectively provide models for volatile reads

Add regression tests for --nondet-volatile-model

cc46483

Add regression tests for invalid usage of --nondet-volatile-model

47ad91a

Add regression test for interaction of volatile/non-volatile variables

d8c6441

danpoe force-pushed the feature/selective-modelling-of-volatile-variables branch 2 times, most recently from 424f44e to d8c6441 Compare June 9, 2020 11:32

danpoe merged commit c7067b0 into diffblue:develop Jun 9, 2020

Selective modelling of volatile variables #5374

Selective modelling of volatile variables #5374

Uh oh!

Conversation

danpoe commented Jun 4, 2020 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

hannes-steffenhagen-diffblue left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

danpoe Jun 9, 2020 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

codecov bot commented Jun 9, 2020 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Codecov Report

Uh oh!

Uh oh!

danpoe commented Jun 4, 2020 •

edited

Loading

danpoe Jun 9, 2020 •

edited

Loading

codecov bot commented Jun 9, 2020 •

edited

Loading