I have setup a project to test the capability of CodeQL，to test taint tracking ablitity

[hatface]

here is my project

https://github.com/hatface/codeql_cap_evalution

here is the test result

https://github.com/hatface/codeql_cap_evalution/blob/main/CodeQL_Capacibility_Test_Result.md

here is the CodeQL Query

https://github.com/hatface/codeql_cap_evalution/blob/main/taint_cap_test.ql

and the case5 is failed

QUESTION： how to fix the situation using isAdditionalTaintStep

[jketema]

Hi @hatface. Thanks for your question. As we have been discussing on https://github.com/github/codeql/issues/11734, CodeQL might not give you back all the paths. When I modify your main function to be just

     source(sourceInt, sourceStr);
     sceneTest05(sourceInt, sourceStr);

the relevant path shows up for me. You really want to disentangle your test cases. So instead of having a single main function with overlapping test cases create 6 separate ones that look something like:

void test0() {
    ...
     source(sourceInt, sourceStr);
     sink(sourceInt, sourceStr);
}

void test1() {
    ...
     source(sourceInt, sourceStr);
     sceneTest01(sourceInt, sourceStr);
}

...

void test5() {
    ...
     source(sourceInt, sourceStr);
     sceneTest05(sourceInt, sourceStr);
}

int main(int argc, char **argv) {
  test0();
  test1();
  ...
  test5();
  return 0;
}

[hatface]

@jketema thx a lot, but I followed your advice, the test case 5 still fail, my project, my QL, my test result updated

[jketema]

Are you sure the path is actually missing? With the latest version of your project I'm seeing the following, which I your 5th test case.

[hatface]

emm， it seems that we encounter a problem，would you pls use the lastest version of my program，and query with the lastest ql。all resource in here https://github.com/hatface/codeql_cap_evalution。 I can't find the testcase 5.

ps：codeql-repo and codeql-binary-cli is the lastest

[jketema]

My result above is with the latest versions.

[jketema]

Which operating system are you on and which compiler do you use?

[hatface]

compile on ubuntu compiler：gcc version 7.5.0 (Ubuntu 7.5.0-3ubuntu1~18.04)

run result on windows with vscode-codeql and codeql on windows

[jketema]

compile on ubuntu compiler：gcc version 7.5.0 (Ubuntu 7.5.0-3ubuntu1~18.04)

I haven't checked on Windows, but I see your issue on Linux. This issue is a combination of the dataflow library you're using and the way shared pointers are implemented by gcc. The solution is to switch to our IR-based dataflow library. To do so, replace

import semmle.code.cpp.dataflow.TaintTracking

by

import semmle.code.cpp.ir.dataflow.TaintTracking

[hatface]

can you share the testcases，which is used by codeQL team？

[jketema]

can you share the testcases，which is used by codeQL team？

I'm not sure what you mean. I just tested with the project you shared.

[hatface]

I mean share CodeQL team's testcases，which your team use to test the ability of CodeQL，or benchmark code，may be

[jketema]

All the test cases are located in https://github.com/github/codeql

[hatface]

compile on ubuntu compiler：gcc version 7.5.0 (Ubuntu 7.5.0-3ubuntu1~18.04)

I haven't checked on Windows, but I see your issue on Linux. This issue is a combination of the dataflow library you're using and the way shared pointers are implemented by gcc. The solution is to switch to our IR-based dataflow library. To do so, replace

import semmle.code.cpp.dataflow.TaintTracking

by

import semmle.code.cpp.ir.dataflow.TaintTracking

by replace this import statement，testcase 5 is passed，but testcase 1, testcase 2, testcase 3, testcase 4 is failed

[hatface]

I haven't checked on Windows

foreign programer used to work on ubuntu，or mac，save a lot of time。 but we programer of china used to work on windows，haha。

[jketema]

by replace this import statement，testcase 5 is passed，but testcase 1, testcase 2, testcase 3, testcase 4 is failed

Try:

from DataFlow::PathNode sink, DataFlow::PathNode source, MyTaintTrackingConfiguration config
where config.hasFlowPath(source, sink)
select sink.getNode().asExpr() ,source, sink,"the taint flow into sink from $@.", source.getNode().asExpr(), "source"

instead of

from DataFlow::PathNode sink, DataFlow::PathNode source, MyTaintTrackingConfiguration config
where config.hasFlowPath(source, sink)
select sink.getNode() ,source, sink,"the taint flow into sink from $@.", source.getNode(), "source"

[hatface]

by replace this import statement，testcase 5 is passed，but testcase 1, testcase 2, testcase 3, testcase 4 is failed

Try:

from DataFlow::PathNode sink, DataFlow::PathNode source, MyTaintTrackingConfiguration config
where config.hasFlowPath(source, sink)
select sink.getNode().asExpr() ,source, sink,"the taint flow into sink from $@.", source.getNode().asExpr(), "source"

instead of

from DataFlow::PathNode sink, DataFlow::PathNode source, MyTaintTrackingConfiguration config
where config.hasFlowPath(source, sink)
select sink.getNode() ,source, sink,"the taint flow into sink from $@.", source.getNode(), "source"

by this，testcase 5,testcase 3,testcase 2，testcase 1 is passed，testcase4 failed

[jketema]

Hi @hatface, I've checked internally and failure of your 4th test case seems to be due to a bug on our side. The bug cannot be worked around with isAdditionalTaintStep. I have filed an internal issue to track this problem.

[ryao]

I haven't checked on Windows

foreign programer used to work on ubuntu，or mac，save a lot of time。 but we programer of china used to work on windows，haha。

This might be useful:

https://ubuntu.com/wsl

It lets you run Ubuntu software on Windows, provided that you have Windows 10 version 1607 or later, and your Windows installation is 64-bit.

[hatface]

@jketema hello jketema，I have new tests，and testcase 14，testcase 17，testcase 18，testcase 21，testcase 22，testcase 23，testcase 24，testcase 30，testcase 33 is failed，the same question how to fix this ，by isAdditionalTaintStep the test result is shown here

https://github.com/hatface/codeql_cap_evalution/blob/main/CodeQL_Capacibility_Test_Result.md

and my project is updated

[jketema]

Hi @hatface,

It's impossible for us to look in detail at your test cases without explanation from your side of (a) what a test case is aimed to test, (b) how you have tried to extend isAdditionalTaintStep yourself to handle a particular test case and why that didn't work, and (c) with which of the data flow libraries the test is failing (semmle.code.cpp.dataflow.TaintTracking, semmle.code.cpp.ir.dataflow.TaintTracking, or both). So, please provide that information.

[hatface]

sorry，I will explain my testcases, the following tables shows the the test case aimed to test.

index	aimed to test	test function	explain	result	other words
14	ternary operator	sceneTest0014	test if when ternary operator in the path is the taint tracking works well	fail	semmle.code.cpp.dataflow.TaintTracking and semmle.code.cpp.ir.dataflow.TaintTracking both fail
17	function pointer	sceneTest0017	test if when function pointor in the path is the taint tracking works well	fail	semmle.code.cpp.dataflow.TaintTracking and semmle.code.cpp.ir.dataflow.TaintTracking both fail
18	function pointer of std::function	sceneTest0018	test if when function pointor of std::function in the path is the taint tracking works well	fail	semmle.code.cpp.dataflow.TaintTracking and semmle.code.cpp.ir.dataflow.TaintTracking both fail
21	virtual function	sceneTest0021	test if when virtual function in the path is the taint tracking works well	fail	semmle.code.cpp.dataflow.TaintTracking and semmle.code.cpp.ir.dataflow.TaintTracking both fail
22	virtual function1	sceneTest0022	test if when virtual function in the path is the taint tracking works well as a positive of 21	fail	semmle.code.cpp.dataflow.TaintTracking and semmle.code.cpp.ir.dataflow.TaintTracking both fail
23	typedefed function pointer	sceneTest0023	test if when typedefed function point in the path is the taint tracking works well	fail	semmle.code.cpp.dataflow.TaintTracking and semmle.code.cpp.ir.dataflow.TaintTracking both fail
24	pre build-in template function	sceneTest0024	test if when pre build-in template function in the path is the taint tracking works well	fail	semmle.code.cpp.dataflow.TaintTracking and semmle.code.cpp.ir.dataflow.TaintTracking both fail
30	static member field	sceneTest0030	test if when static member field in the path is the taint tracking works well	fail	semmle.code.cpp.dataflow.TaintTracking and semmle.code.cpp.ir.dataflow.TaintTracking both fail
31	decontruction function using with smart pointer	sceneTest0033	test if whendecontruction function using with smart pointer in the path is the taint tracking works well	fail	semmle.code.cpp.dataflow.TaintTracking and semmle.code.cpp.ir.dataflow.TaintTracking both fail

[jketema]

I will explain my testcases, the following tables shows the the test case aimed to test.

Thanks. So in each case how did you try to extend isAdditionalTaintStep to try to fix the problem?

[hatface]

I did'nt extend isAdditionalTaintStep by myself, but some of the test case I have tried before, for example testcase 30, I can't find an elegent way to extend isAdditionalTaintStep in the way of programing language, but only extend isAdditionalTaintStep in the way of business, so, I test the programing language feature, and marked all the failed testcase, as a first step, lying them in the issue, and looking for help. I do want to fix them case by case.

[jketema]

I'm happy to help, but do expect some kind of effort from your side. In this case you should first really try to modify isAdditionalTaint step yourself. We cannot do all your work for you.