[lldb-dev] raw c strings in lldb

[Reid Kleckner]

I'd also point out that if you hate writing "llvm::" or using decls
everywhere, you can import StringRef into the lldb namespace. Clang does
this in include/clang/Basic/LLVM.h for StringRef and other common classes.

On Wed, Feb 11, 2015 at 10:18 AM, Zachary Turner <zturner at google.com> wrote:

> A patch is up <http://reviews.llvm.org/D7553> which fixes a number of
> issues with strings not being null terminated in LLDB.  The issues and
> error-proneness of using raw c strings is well documented and understood,
> so I would like to propose banning them in LLDB for new code [1].
>
> I realize that's a tall order, and I don't expect all the occurrences of
> them to go away overnight, but issues related to strings have been popping
> up over and over again, and it would be nice if we could deal with more
> interesting problems.
>
> LLVM has a very robust string manipulation library, and while it's not
> perfect, it would have prevented almost all of these issues from ever
> happening in the first place.  Here is a quick summary of common C string
> types / operations and their corresponding LLVM equivalents:
>
> Types:
> char stack_buffer[n];        // Use llvm::SmallString<n>.
> const char* foo;             // Use llvm::StringRef foo.
>
> Passing arguments;
> void foo(char *, int len);   // Use void foo(llvm::SmallVectorImpl<char>
> &buf);
> void foo(const char * foo);  // Use void foo(llvm::StringRef foo);
>
> Operations:
> strcpy, strncpy, etc         // Use member functions of SmallString /
> StringRef
> sprintf, snprintf, etc       // Use llvm::raw_ostream
>
> We can fix existing occurrences gradually / as time permits, but I would
> prefer to see a trend towards new code using llvm's string classes and
> formatting library, and when you are touching a piece of code that is
> messing with raw c-strings, that's a perfect time to change that code to
> migrate to LLVM strings.
>
> [1] There are a couple of places where we can't ban them entirely.  This
> relates to the public API (i.e. can't change an existing signature), and
> va_args off the top of my head.  The va_args stuff though, if generally
> useful, if stuff we can sink into LLVM, there's nothing debuggery about
> formatting strings.
>
> Thoughts?  Suggestions?  Comments?
>
> _______________________________________________
> lldb-dev mailing list
> lldb-dev at cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/lldb-dev
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/lldb-dev/attachments/20150211/d5913cc8/attachment.html>