[llvm-dev] [Proposal][Debuginfo] dsymutil-like tool for ELF.

Alexey via llvm-dev llvm-dev at lists.llvm.org
Tue Aug 25 07:29:18 PDT 2020


   We propose llvm-dwarfutil - a dsymutil-like tool for ELF.
   Any thoughts on this?
   Thanks in advance, Alexey.


llvm-dwarfutil(Apndx A) - is a tool that is used for processing debug 
located in built binary files to improve debug info quality,
reduce debug info size and accelerate debug info processing.
Supported object files formats: ELF, MachO(Apndx B), COFF(Apndx C), 
WASM(Apndx C).


Specifically, the tool would do:

   - Remove obsolete debug info which refers to code deleted by the linker
     doing the garbage collection (gc-sections).

   - Deduplicate debug type definitions for reducing resulting size of 

   - Build accelerator/index tables.
     = .debug_aranges, .debug_names, .gdb_index, .debug_pubnames, 

   - Strip unneeded tables.
     = .debug_aranges, .debug_names, .gdb_index, .debug_pubnames, 

   - Compress or decompress debug info as requested.

Possible feature:

   - Join split dwarf .dwo files in a single file containing all debug info
     (convert split DWARF into monolithic DWARF).


User interface:

   OVERVIEW: A tool for optimizing debug info located in the built binary.

   USAGE: llvm-dwarfutil [options] input output

   OPTIONS: (Apndx E)


Implementation notes:

1. Removing obsolete debug info would be done using DWARFLinker llvm 

2. Data types deduplication would be done using DWARFLinker llvm library.

3. Accelerator/index tables would be generated using DWARFLinker llvm 

4. Interface of DWARFLinker library would be changed in such way that it
    would be possible to switch on/off various stages:

   class DWARFLinker {
     setDoRemoveObsoleteInfo ( bool DoRemoveObsoleteInfo = false);

     setDoAppleNames ( bool DoAppleNames = false );
     setDoAppleNamespaces ( bool DoAppleNamespaces = false );
     setDoAppleTypes ( bool DoAppleTypes = false );
     setDoObjC ( bool DoObjC = false );
     setDoDebugPubNames ( bool DoDebugPubNames = false );
     setDoDebugPubTypes ( bool DoDebugPubTypes = false );

     setDoDebugNames (bool DoDebugNames = false);
     setDoGDBIndex (bool DoGDBIndex = false);

5. Copying source file contents, stripping tables, 
compressing/decompressing tables
    would be done by ObjCopy llvm library(extracted from llvm-objcopy):

   Error executeObjcopyOnBinary(const CopyConfig &Config,
                              object::COFFObjectFile &In, Buffer &Out);
   Error executeObjcopyOnBinary(const CopyConfig &Config,
                              object::ELFObjectFileBase &In, Buffer &Out);
   Error executeObjcopyOnBinary(const CopyConfig &Config,
                              object::MachOObjectFile &In, Buffer &Out);
   Error executeObjcopyOnBinary(const CopyConfig &Config,
                              object::WasmObjectFile &In, Buffer &Out);

6. Address ranges and single addresses pointing to removed code should 
be marked
    with tombstone value in the input file:

    -2 for .debug_ranges and .debug_loc.
    -1 for other .debug* tables.

7. Prototype implementation - https://reviews.llvm.org/D86539.



1. Refactor llvm-objcopy to extract it`s implementation into separate 
    ObjCopy(in LLVM tree).

2. Create a command line utility using existed DWARFLinker and ObjCopy
    implementation. First version is supposed to work with only ELF 
input object files.
    It would take input ELF file with unoptimized debug info and create 
    ELF file with optimized debug info. That version would be done out 
of the llvm tree.

3. Make a tool to be able to work in multi-thread mode.

4. Consider it to be included into LLVM tree.

5. Support DWARF5 tables.


Appendix A. Should this tool be implemented as a new tool or as an extension
             to dsymutil/llvm-objcopy?

    There already exists a tool which removes obsolete debug info on 
darwin - dsymutil.
    Why create another tool instead of extending the already existed 

    The main functionality of dsymutil is located in a separate library 
- DWARFLinker.
    Thus, dsymutil utility is a command-line interface for DWARFLinker. 
dsymutil has
    another type of input/output data: it takes several object files and 
address map
    as input and creates a .dSYM bundle with linked debug info as 
output. llvm-dwarfutil
    would take a built executable as input and create an optimized 
executable as output.
    Additionally, there would be many command-line options specific for 
only one utility.
    This means that these utilities(implementing command line interface) 
would significantly
    differ. It makes sense not to put another command-line utility 
inside existing dsymutil,
    but make it as a separate utility. That is the reason why 
llvm-dwarfutil suggested to be
    implemented not as sub-part of dsymutil but as a separate tool.

    Please share your preference: whether llvm-dwarfutil should be
    separate utility, or a variant of dsymutil compiled for ELF?


Appendix B. The machO object file format is already supported by dsymutil.
    Depending on the decision whether llvm-dwarfutil would be done as a 
    of dsymutil or as a separate utility - machO would be supported or not.


Appendix C. Support for the COFF and WASM object file formats presented as
     possible future improvement. It would be quite easy to add them 
     that llvm-objcopy already supports these formats. It also would require
     supporting DWARF6-suggested tombstone values(-1/-2).


Appendix D. Documentation.

   - proposal for DWARF6 which suggested -1/-2 values for marking bad 
   - dsymutil tool https://llvm.org/docs/CommandGuide/dsymutil.html.
   - proposal "Remove obsolete debug info in lld."


Appendix E. Possible command line options:

DwarfUtil Options:

   --build-aranges           - generate .debug_aranges table.
   --build-debug-names       - generate .debug_names table.
   --build-debug-pubnames    - generate .debug_pubnames table.
   --build-debug-pubtypes    - generate .debug_pubtypes table.
   --build-gdb-index         - generate .gdb_index table.
   --compress                - Compress debug tables.
   --decompress              - Decompress debug tables.
   --deduplicate-types       - Do ODR deduplication for debug types.
   --garbage-collect         - Do garbage collecting for debug info.
   --num-threads=<n>         - Specify the maximum number (n) of 
simultaneous threads
                               to use when optimizing input file.
                               Defaults to the number of cores on the 
current machine.
   --strip-all               - Strip all debug tables.
   --strip=<name1,name2>     - Strip specified debug info tables.
   --strip-unoptimized-debug - Strip all unoptimized debug tables.
   --tombstone=<value>       - Tombstone value used as a marker of 
invalid address.
     =bfd                    -   BFD default value
     =dwarf6                 -   Dwarf v6.
   --verbose                 - Enable verbose logging and encoding details.

Generic Options:

   --help                    - Display available options (--help-hidden 
for more)
   --version                 - Display the version of this program

More information about the llvm-dev mailing list