This basically jsut pulls the major components of the module debug info stream out. It doesn't interpret any of this data semantically because that requires more investigation. But hopefully this at least presents a high level picture of what data is in each module's debug info stream. Subsequent patches will dig deeper and try to parse the symbol substream and the line info substreams.
Details
Diff Detail
Event Timeline
All of this code is still dead. Are you sure you don't want to hold off on committing this until some kind of byte dumping is hooked up from llvm-pdbdump?
include/llvm/DebugInfo/PDB/Raw/ModStream.h | ||
---|---|---|
11 | Wouldn't the header guard be named ...RAW_MODSTREAM_H? | |
lib/DebugInfo/PDB/CMakeLists.txt | ||
36 | consider sorting these at some point | |
lib/DebugInfo/PDB/Raw/ModStream.cpp | ||
32 | Same thought as before about adding errorCorruptFile or something. |
lib/DebugInfo/PDB/Raw/ByteStream.cpp | ||
---|---|---|
27–29 | Why do you manage ownership by hand instead of unique_ptr? |
lib/DebugInfo/PDB/Raw/ByteStream.cpp | ||
---|---|---|
27–29 | Whether or not the data is owned should be optional. I could change the bool to a unique_ptr and only initialize it optionally, seems like that would work and be fine. |
Updated based on suggestions. Also now dumps some information from each module's stream, so that it's not dead code anymore.
Wouldn't the header guard be named ...RAW_MODSTREAM_H?