Introduce shared taint tracking library #13881
Conversation
|
A few stylistic comments, otherwise mostly LGTM. |
| private import DataFlowMake<DataFlowLang> as DataFlow | ||
| private import MakeImpl<DataFlowLang> as DataFlowInternal | ||
|
|
||
| private module AddTaintDefaults<DataFlowInternal::FullStateConfigSig Config> implements |
Check warning
Code scanning / CodeQL
Data flow configuration module naming Warning
| * Constructs a global taint tracking computation. | ||
| */ | ||
| module Global<DataFlow::ConfigSig Config> implements DataFlow::GlobalFlowSig { | ||
| private module Config0 implements DataFlowInternal::FullStateConfigSig { |
Check warning
Code scanning / CodeQL
Data flow configuration module naming Warning
| import Config | ||
| } | ||
|
|
||
| private module C implements DataFlowInternal::FullStateConfigSig { |
Check warning
Code scanning / CodeQL
Data flow configuration module naming Warning
| * Constructs a global taint tracking computation using flow state. | ||
| */ | ||
| module GlobalWithState<DataFlow::StateConfigSig Config> implements DataFlow::GlobalFlowSig { | ||
| private module Config0 implements DataFlowInternal::FullStateConfigSig { |
Check warning
Code scanning / CodeQL
Data flow configuration module naming Warning
| import Config | ||
| } | ||
|
|
||
| private module C implements DataFlowInternal::FullStateConfigSig { |
Check warning
Code scanning / CodeQL
Data flow configuration module naming Warning
|
LGTM now. Although, preferably we'll switch all languages in one go, so we don't have a mix of shared lib usage and parameterised modules. |
That is the plan (as part of this PR). I only did C++ first, because I wanted to know if the approach was good. |
|
Beware that you have an incoming merge conflict in #13851 |
I'll rebase this PR and then kick off all the DCAs. |
9f12a4e
to
c4a65e5
Compare
|
I sightly refactored this to follow #13901 |
|
@aschackmull I'm slightly confused by the DCA results. I see OOMs in Java (which seem to be sort of expected), C# and C++. I tried the C++ and C# OOM'ing projects on larger runners to see if there are any suspicious tuple counts stage timings. I see slightly slower stage timings, in a few cases but nothing really jumps out to me. |
No description provided.