chore(deps): update rust crate iai-callgrind to 0.14

[renovate]

This PR contains the following updates:

Package	Type	Update	Change
iai-callgrind	workspace.dependencies	minor	0.13 -> 0.14

---

Release Notes

iai-callgrind/iai-callgrind (iai-callgrind)

v0.14.0

Compare Source

This release adds full support for multi-threaded and multi-process
applications.

When upgrading from a previous release of Iai-Callgrind you might experience
changes in the metrics without having changed the benchmarks themselves. The
summary line in callgrind output files turned out to be buggy and unreliable
if client requests are used, so Iai-Callgrind now parses the totals instead.
The totals might differ slightly from the summary and cause the difference
in the displayed metrics. You might also see changes in the metrics because of
the changed default values for some of the valgrind arguments. Iai-Callgrind
changed the following default valgrind/callgrind arguments for each benchmark
run:

• --separate-threads=no -> --separate-threads=yes
• --trace-children=no -> --trace-children=yes
• --fair-sched=no -> --fair-sched=try

You can now run the setup in binary benchmarks in parallel to the Command
for flexible benchmarking of client/server architectures.

The MSRV has changed from 1.66.0 -> 1.67.1.

If not stated otherwise the changes below were introduced in
#​263.

Added

• Support for benchmarks of multi-threading and multi-process applications by
  implementing the correct handling of the valgrind --trace-children and
  callgrind --separate-threads command line options. Per default only the
  total over all subprocesses and threads is calculated and shown. But, each
  thread and subprocess can be displayed with the new
  OutputFormat::show_intermediate option.
• Support for the callgrind command line arguments --dump-every-bb,
  --dump-before, --dump-after which create parts. These parts are now
  correctly summarized in the total and the metrics of each part can be shown
  with OutputFormat::show_intermediate.
• Added OutputFormat which can be used in
  LibraryBenchmarkConfig::output_format and
  BinaryBenchmarkConfig::output_format to change some of the behaviour of the
  default terminal output (but not json output)
• Sometimes callgrind creates empty files, so we're cleaning them up now after
  each benchmark run.
• (#​256) and
  (#​279): Support
  running setup in parallel and add possibility to delay the Command. Thanks
  to @​hargut for #​256
• (#​288): Added an
  option OutputFormat::show_grid to show grid/guiding lines which can help
  reading the terminal output if running benchmarks with multiple
  threads/subprocesses/tools.
• The method BinaryBenchmarkConfig::with_callgrind_args was added to match the
  constructors of the LibraryBenchmarkConfig.
• The methods BinaryBenchmarkConfig::valgrind_args and
  LibraryBenchmarkConfig::valgrind_args are introduced to be able to pass
  valgrind core arguments to all tools.

Changed

• All tools are now per default run with --trace-children=yes and
  --fair-sched=try. In addition, callgrind is run with
  --separate-threads=yes. These default arguments can be changed in
  Tool::args or LibraryBenchmarkConfig::callgrind_args,
  BinaryBenchmarkConfig::callgrind_args.

• The file naming scheme was adjusted to include the pids in case of
  multi-process benchmarks, the parts in case of callgrind command-line
  arguments which create multiple parts and threads in case of multiple threads.
  This change is backwards compatible to the file naming scheme of previous
  Iai-Callgrind releases for all tools but exp-bbv.

• Error metrics from tools like drd, helgrind and memcheck are now listed and
  compared like the other metrics in a vertical format. For example

  ======= DRD ===============================================================
  Errors:                           0|0               (No change)
  Contexts:                         0|0               (No change)
  Suppressed Errors:                0|0               (No change)
  Suppressed Contexts:              0|0               (No change)
  

• (#​263) and
  (#​288): Increase
  the field width by 3 bytes and the space for metrics by 5 on each side of the
  comparison so that the value of u64::MAX fits into the terminal output
  without messing up the side-by-side layout.

• The LibraryBenchmarkConfig::truncate_description,
  BinaryBenchmarkConfig::truncate_description methods have been moved to
  OutputFormat::truncate_description

• In the presence of multiple processes the DHAT metrics are now summarized and
  shown in a total in the same way as the metrics of callgrind and the other
  tools.

• Bump the summary json schema to v3 in
  iai-callgrind-runner/schemas/summary.v3.schema.json

• Various prs: Update locked direct dependencies:

  • anyhow -> 1.0.89
  • cc -> 1.1.25
  • indexmap -> 2.6.0
  • itertools -> 0.13.0
  • once_cell -> 1.20.1
  • regex -> 1.11.0
  • serde_json -> 1.0.128
  • serde -> 1.0.210
  • syn -> 2.0.79
  • tempfile -> 3.13.0

• (#​288): The default
  include path for the valgrind headers has changed to /usr/local/include on
  freebsd instead of /usr/local.

• (#​289): Update
  derive_more -> 1.0 in Cargo.toml but not in lock file.

• (#​293): Update MSRV
  from 1.66.0 -> 1.67.1

• (#​296): Update
  locked transitive dependencies.

Deprecated

• The following methods were renamed and deprecate the old method name:
  • LibraryBenchmarkConfig::raw_callgrind_args ->
    LibraryBenchmarkConfig::callgrind_args,
  • LibraryBenchmarkConfig::with_raw_callgrind_args ->
    LibraryBenchmarkConfig::with_callgrind_args
  • BinaryBenchmarkConfig::raw_callgrind_args ->
    BinaryBenchmarkConfig::callgrind_args
• The method LibraryBenchmarkConfig::raw_callgrind_args_iter was deprecated
  since it is the same as LibraryBenchmarkConfig::callgrind_args.

Removed

• Iai-Callgrind doesn't support combined dumps via --combine-dumps anymore.
• The Tool::outfile_modifier method was removed. The %p modifier for
  valgrind output and log files is now applied automatically when using the
  --trace-children=yes command line argument.
• The output and log file paths in the terminal output were removed.

Fixed

• When extracting the metrics from callgrind output files, the totals line is
  now prioritized over the summary line. The summary line has bugs and reports
  wrong costs if callgrind client requests are used. The totals are unaffected
  by client requests and report the correct costs. This change is mostly
  internal but might introduce some (small) changes in the metrics reported by
  Iai-Callgrind.
• The error metrics of drd, helgrind and memcheck were only shown correctly if
  they consisted of a single digit.
• (#​297): Added the
  derive Clone impl for iai_callgrind::LibraryBenchmarkConfig
• (#​300):
  library_benchmark_group! was private but the expanded mod should be public
  Thanks to @​ryanpeach

v0.13.4

Compare Source

Changed

• (#​264): Migrate
  from unmaintained proc-macro-error to proc-macro-error2 due to
  https://rustsec.org/advisories/RUSTSEC-2024-0370. This also removes the
  duplicate dependency on syn v2 and syn v1.

---

Configuration

📅 Schedule: Branch creation - At any time (no schedule defined), Automerge - At any time (no schedule defined).

🚦 Automerge: Disabled by config. Please merge this manually once you are satisfied.

♻ Rebasing: Whenever PR becomes conflicted, or you tick the rebase/retry checkbox.

🔕 Ignore: Close this PR and you won't be reminded about this update again.

---

• If you want to rebase/retry this PR, check this box

---

This PR was generated by Mend Renovate. View the repository job log.