[PATCH] D30341: [analyzer] clarify error messages about uninitialized function arguments

[Daniel Marjamäki via Phabricator via cfe-commits]

danielmarjamaki added inline comments.


================
Comment at: lib/StaticAnalyzer/Checkers/CallAndMessageChecker.cpp:211
       // Generate a report for this bug.
-      StringRef Desc =
-          describeUninitializedArgumentInCall(Call, IsFirstArgument);
+      std::string Desc =
+          describeUninitializedArgumentInCall(Call, ArgumentNumber);
----------------
danielmarjamaki wrote:
> zaks.anna wrote:
> > danielmarjamaki wrote:
> > > zaks.anna wrote:
> > > > Have you considered using  llvm::raw_svector_ostream here as well as passing it an argument to describeUninitializedArgumentInCall? For example, see  MallocChecker.cpp.
> > > I changed so describeUninitializedArgumentInCall() returns an llvm::Twine instead of std::string. hope you like it.
> > > 
> > I do not think it's safe to use llvm:Twine here. See http://llvm.org/docs/ProgrammersManual.html#the-twine-class
> > 
> > How about using llvm::raw_svector_ostream as I suggested?
> sure I can use llvm::raw_svector_ostream instead. I can try to update the patch soon.
> 
> I just wonder how it is unsafe. I did consider if llvm::Twine would be safe. Is there a particular return that you can point out? The function mostly returns constant string literals. Those should be safe right? Then the function also have a few returns like this:
> ```
>     return llvm::Twine(ArgumentNumber + 1) +
>                llvm::getOrdinalSuffix(ArgumentNumber + 1) +
>                " function call argument is an uninitialized value";
> ```
> Yes we need to be careful for such code. However since llvm::getOrdinalSuffix() returns a StringRef this particular code should be safe right?
> 
> When the stream is used I have to tweak each return statement. I thought that was a bit unfortunate. But it's not a biggie.
> 
After some more reading in the llvm::Twine docs I am also thinking that I misuse it. so I will change the code.


Repository:
  rL LLVM

https://reviews.llvm.org/D30341