lcov2xml

Executable to convert lcov info files to cobertura XML format.

A more performant standalone application replacing the Python based lcov-to-cobertura-xml

Contains two tools to help with CI/CD coverage, especially for GitLab. lcov2xml to convert and if needed cobertura_split to chunk them into allowed sizes for GitLab artifacts.

You can avoid installing those tools, because conversion functionality is also integrated in cargo-llvm-cov which might already covers your needs.

Features

• can demangle C++ names
• can demangle rustc names
• merges multiple lcov reports into one
• can split big XML files into many smaller ones for GitLab attachment size limitation. strategy: it generates 9.5MB big xml files, fitting as many packages as possible into each file

Usage

# install using cargo
cargo install lcov2xml
# inspect usage
lcov2xml --help
# this would write file coverage.xml
lcov2xml lcov.info
# this splits an existing xml file into smaller ones
cobertura_split coverage.xml

available cmd-line args

convert LCOV info file to cobertura XML format

Usage: lcov2xml [OPTIONS] [FILES]...

Arguments:
  [FILES]...  LCOV input files, use single dash '-' argument to read from stdin

Options:
  -b, --base-dir <BASE_DIR>    Directory where source files are located [default: .]
  -o, --output <OUTPUT>        Path to store cobertura xml file [default: coverage.xml]
  -e, --excludes <EXCLUDES>    Comma-separated list of regexes of packages to exclude [default: ]
  -d, --demangle               Demangle function names
      --demangler <DEMANGLER>  Path to demangler tool, e.g. c++filt for C++, $rust = internal rustc demangler [default: $rust]
      --split-xml              splits XML file into 9.5MB big chunks for GitLab, attention keeps original file intact
  -h, --help                   Print help information
  -V, --version                Print version information

Performance

Ran on a about 500KiB sized lcov.info file on macOS and measured the wall-clock time plus max RAM usage. Is faster in gross runtime but more importantly uses an order of magnitude less RAM.

For the coverage.xml splitting tool input is a 100MB sized xml file. RAM usage is drastically reduced, so is runtime.

All times are measured after some warm-up runs to fill the disk cache.

/usr/bin/time -al python3 lcov_cobertura.py lcov.info
/usr/bin/time -al target/release/lcov2xml lcov.info

/usr/bin/time -al python3 split-by-package-int.py huge.xml outdir
/usr/bin/time -al target/release/cobertura_split huge.xml