[LLVMbugs] [Bug 184] NEW: [bcreader/writer] include/llvm/Bytecode/Primitives.h should be cleaned up

[bugzilla-daemon at zion.cs.uiuc.edu]

http://llvm.cs.uiuc.edu/bugs/show_bug.cgi?id=184

           Summary: [bcreader/writer] include/llvm/Bytecode/Primitives.h
                    should be cleaned up
           Product: libraries
           Version: 1.0
          Platform: All
        OS/Version: All
            Status: NEW
          Severity: minor
          Priority: P2
         Component: Bytecode Reader
        AssignedTo: sabre at nondot.org
        ReportedBy: sabre at nondot.org


include/llvm/Bytecode/Primitives.h is a gross hack that desperately needs to be
cleaned up.  In particular, at least the following should be done:

1. The file should be split into two distinct pieces (one for the reader,
   one for the writer) and the pieces should be moved into lib/Bytecode/*

2. The reader stuff should throw exceptions on errors, so that the bytecode
   reader doesn't have to transform the true/false return codes into exceptions.
   This will likely also make the code more efficient, and making the bytecode
   reader as efficient as possible is a high priority for LLVM.

3. Using a deque<unsigned char> to accumulate the bytecode file in the writer
   is pretty gross, and certainly has to be inefficient.  A custom container
   should be implemented for this, because we really only need "append with
   backpatching" support.

4. Instead of conditionally compiling the code based on endianness, it seems
   to make sense to use the network byte order functions to do byteswapping
   for us.  The only problem is that LLVM bytecode files are not in network
   byte order (big endian) they are little endian.

-Chris



------- You are receiving this mail because: -------
You are on the CC list for the bug, or are watching someone who is.