As an aside, I'm actually currently working on a tool that parses the Linux source code to find symbol definitions and then works its way back through the Git history to find the commit message for when the symbol was first defined.
Are you just doing something similar to cscope to find the definition of a symbol, then running git blame on that line? Or are you actually checking earlier revisions as well, to see if the symbol was moved or changed types?
I'm finding the definition of a symbol using a parsing expression grammar for C. I haven't yet decided upon the particular algorithm for finding the first commit, but I plan on using libgit2 to work directly with the repository. In the simplest case, you just iterate through all versions of a file and find the earliest one with the string in question. Obviously care is required when walking the repository history.
Are you just doing something similar to cscope to find the definition of a symbol, then running git blame on that line? Or are you actually checking earlier revisions as well, to see if the symbol was moved or changed types?