v7.13.4

Compare Source

• Fix: the third-party code fix in 7.13.3 required examining the parent
  directories where coverage was run. In the unusual situation that one of the
  parent directories is unreadable, a PermissionError would occur, as
  described in issue 2129_. This is now fixed.

• Fix: in test suites that change sys.path, coverage.py could fail with
  "RuntimeError: Set changed size during iteration" as described and fixed in
  pull 2130_. Thanks, Noah Fatsi.

• We now publish ppc64le wheels, thanks to Pankhudi Jain _.

.. _pull 2121: # 2121
.. _issue 2129: # 2129
.. _pull 2130: # 2130

.. _changes_7-13-3:

v7.13.3

Compare Source

• Fix: in some situations, third-party code was measured when it shouldn't have
  been, slowing down test execution. This happened with layered virtual
  environments such as uv sometimes makes. The problem is fixed, closing issue 2082_. Now any directory on sys.path that is inside a virtualenv is
  considered third-party code.

.. _issue 2082: # 2082

.. _changes_7-13-2:

v7.13.2

Compare Source

• Fix: when Python is installed via symlinks, for example with Homebrew, the
  standard library files could be incorrectly included in coverage reports.
  This is now fixed, closing issue 2115_.

• Fix: if a data file is created with no read permissions, the combine step
  would fail completely. Now a warning is issued and the file is skipped.
  Closes issue 2117_.

.. _issue 2115: # 2115
.. _issue 2117: # 2117

.. _changes_7-13-1:

v7.13.1

Compare Source

• Added: the JSON report now includes a "start_line" key for function and
  class regions, indicating the first line of the region in the source. Closes
  issue 2110_.

• Added: The debug data command now takes file names as arguments on the
  command line, so you can inspect specific data files without needing to set
  the COVERAGE_FILE environment variable.

• Fix: the JSON report used to report module docstrings as executed lines,
  which no other report did, as described in issue 2105_. This is now fixed,
  thanks to Jianrong Zhao.

• Fix: coverage.py uses a more disciplined approach to detecting where
  third-party code is installed, and avoids measuring it. This shouldn't change
  any behavior. If you find that it does, please get in touch.

• Performance: data files that will be combined now record their hash as part
  of the file name. This lets us skip duplicate data more quickly, speeding the
  combining step.

• Docs: added a section explaining more about what is considered a missing
  branch and how it is reported: :ref:branch_explain, as requested in issue 1597. Thanks to Ayisha Mohammed .

• Tests: the test suite misunderstood what core was being tested if
  COVERAGE_CORE wasn't set on 3.14+. This is now fixed, closing issue 2109_.

.. _issue 1597: # 1597
.. _pull 2092: # 2092
.. _issue 2105: # 2105
.. _issue 2109: # 2109
.. _issue 2110: # 2110

.. _changes_7-13-0:

v7.13.0

Compare Source

• Feature: coverage.py now supports :file:.coveragerc.toml configuration
  files. These files use TOML syntax and take priority over
  :file:pyproject.toml but lower priority than :file:.coveragerc files.
  Closes issue 1643_ thanks to Olena Yefymenko _.

• Fix: we now include a permanent .pth file which is installed with the code,
  fixing issue 2084. In 7.12.1b1 this was done incorrectly: it didn't work
  when using the source wheel (py3-none-any). This is now fixed. Thanks,
  Henry Schreiner .

• Deprecated: when coverage.py is installed, it creates three command entry
  points: coverage, coverage3, and coverage-3.10 (if installed for
  Python 3.10). The second and third of these are not needed and will
  eventually be removed. They still work for now, but print a message about
  their deprecation.

.. _issue 1643: # 1643
.. _pull 1952: # 1952
.. _pull 2100: # 2100

.. _changes_7-12-1b1:

v7.12.0

Compare Source

• The HTML report now shows separate coverage totals for statements and
  branches, as well as the usual combined coverage percentage. Thanks to Ryuta
  Otsuka for the discussion _ and the implementation _.

• The JSON report now includes separate coverage totals for statements and
  branches, thanks to Ryuta Otsuka _.

• Fix: except* clauses were not handled properly under the "sysmon"
  measurement core, causing KeyError exceptions as described in issue 2086_.
  This is now fixed.

• Fix: we now defend against aggressive mocking of open() that could cause
  errors inside coverage.py. An example of a failure is in issue 2083_.

• Fix: in unusual cases where a test suite intentionally exhausts the system's
  file descriptors to test handling errors in open(), coverage.py would
  fail when trying to open source files, as described in issue 2091_. This
  is now fixed.

• A small tweak to the HTML report: file paths now use thin spaces around
  slashes to make them easier to read.

.. _issue 2081: # 2081
.. _issue 2083: # 2083
.. _pull 2085: # 2085
.. _issue 2086: # 2086
.. _pull 2090: # 2090
.. _issue 2091: # 2091

.. _changes_7-11-3:

v7.11.3

Compare Source

• Fix: the 7.11.1 changes meant that conflicts between a requested measurement
  core and other settings would raise an error. This was a breaking change from
  previous behavior, as reported in issue 2076_ and issue 2078_.

  The previous behavior has been restored: when the requested core conflicts
  with other settings, another core is used instead, and a warning is issued.

• For contributors: the repo has moved from Ned's nedbat GitHub account_ to
  the coveragepy GitHub organization_. The default branch has changed from
  master to main.

.. _issue 2076: # 2076
.. _issue 2078: # 2078
.. _nedbat GitHub account: https://github.com/nedbat
.. _coveragepy GitHub organization: https://github.com/coveragepy

.. _changes_7-11-2:

v7.11.2

Compare Source

• Fix: using the "sysmon" measurement core in 7.11.1, if Python code was
  claimed to come from a non-Python file, a NotPython exception could be
  raised. This could happen for example with Jinja templates compiled to
  Python, as reported in issue 2077_. This is now fixed.

• Doc: corrected the first entry in the 7.11.1 changelog.

.. _issue 2077: # 2077

.. _changes_7-11-1:

v7.11.1

Compare Source

• Fix: some chanages to details of how the measurement core is chosen, and how
  conflicting settings are handled. The "sysmon" core cannot be used with some
  conurrency settings, with dynamic context, and in Python 3.12/3.13, with
  branch measurement.

  • If the core is not specified and defaults to "sysmon" (Python 3.14+), but
    other settings conflict with sysmon, then the "ctrace" core will be used
    instead with no warning. For concurrency conflicts, this used to produce an
    error, as described in issue 2064_.

  • If the "sysmon" core is explicitly requested in your configuration, but
    other settings conflict, an error is now raised. This used to produce a
    warning.

• Fix: some multi-line case clauses or for loops (and probably other
  constructs) could cause incorrect claims of missing branches with the
  sys.monitoring core, as described in issue 2070_. This is now fixed.

• Fix: when running in pytest under coverage, a breakpoint() would stop in
  the wrong frame, one level down from where it should, as described in issue 1420_. This was due to a coverage change in v6.4.1 that seemed to give a
  slight performance improvement, but I couldn't reproduce the performance
  gain, so it's been reverted, fixing the debugger problem.

• A new debug option --debug=core shows which core is in use and why.

• Split sqlite debugging information out of the sys :ref:coverage debug and :ref:cmd_run_debug options since it's bulky and not
  very useful.

• Updated the :ref:howitworks page to better describe the three different
  measurement cores.

.. _issue 1420: # 1420
.. _issue 2064: # 2064
.. _issue 2070: # 2070

.. _changes_7-11-0:

v7.11.0

Compare Source

• Dropped support for Python 3.9, declared support for Python 3.15 alpha.

.. _changes_7-10-7:

v7.10.7

Compare Source

• Performance: with branch coverage in large files, generating HTML, JSON, or
  LCOV reports could take far too long due to some quadratic behavior when
  creating the function and class index pages. This is now fixed, closing
  issue 2048_. Thanks to Daniel Diniz for help diagnosing the problem.

• Most warnings and a few errors now have links to a page in the docs
  explaining the specific message. Closes issue 1921_.

.. _issue 1921: # 1921
.. _issue 2048: # 2048

.. _changes_7-10-6:

v7.10.6

Compare Source

• Fix: source directories were not properly communicated to subprocesses
  that ran in different directories, as reported in issue 1499_. This is now
  fixed.

• Performance: Alex Gaynor continues fine-tuning _ the speed of
  combination, especially with many contexts.

.. _issue 1499: # 1499
.. _pull 2038: # 2038

.. _changes_7-10-5:

v7.10.5

Compare Source

• Big speed improvements for coverage combine: it's now about twice as
  fast! Huge thanks to Alex Gaynor for pull requests 2032 ,
  2033 , and 2034 _.

.. _pull 2032: # 2032
.. _pull 2033: # 2033
.. _pull 2034: # 2034

.. _changes_7-10-4:

v7.10.4

Compare Source

• Added patch = fork for times when the built-in forking support is
  insufficient.

• Fix: patch = execv also inherits the entire coverage configuration now.

.. _changes_7-10-3:

v7.10.3

Compare Source

• Fixes for patch = subprocess:

  • If subprocesses spawned yet more subprocesses simultaneously, some coverage
    could be missed. This is now fixed, closing issue 2024_.

  • If subprocesses were created in other directories, their data files were
    stranded there and not combined into the totals, as described in issue 2025_. This is now fixed.

  • On Windows (or maybe only some Windows?) the patch would fail with a
    ModuleNotFound error trying to import coverage. This is now fixed,
    closing issue 2022_.

  • Originally only options set in the coverage configuration file would apply
    to subprocesses. Options set on the coverage run command line (such as
    --branch) wouldn't be communicated to the subprocesses. This could
    lead to combining failures, as described in issue 2021_. Now the entire
    configuration is used in subprocesses, regardless of its origin.

  • Added debug=patch to help diagnose problems.

• Fix: really close all SQLite databases, even in-memory ones. Closes issue 2017_.

.. _issue 2017: # 2017
.. _issue 2021: # 2021
.. _issue 2022: # 2022
.. _issue 2024: # 2024
.. _issue 2025: # 2025

.. _changes_7-10-2:

v7.10.2

Compare Source

• Fix: some code with NOP bytecodes could report missing branches that are
  actually executed. This is now fixed, closing issue 1999_. Python 3.9
  still shows the problem.

.. _issue 1999: # 1999

.. _changes_7-10-1:

v7.10.1

Compare Source

• Fix: the exclusion for if TYPE_CHECKING: was wrong: it marked the branch
  as partial, but it should have been a line exclusion so the entire clause
  would be excluded. Improves issue 831_.

• Fix: changed where .pth files are written for patch = subprocess, closing
  issue 2006_.

.. _issue 2006: # 2006

.. _changes_7-10-0:

v7.10.0

Compare Source

• A new configuration option: ":ref:config_run_patch" specifies named patches
  to work around some limitations in coverage measurement. These patches are
  available:

  • patch = _exit lets coverage save its data even when :func:os._exit() is used to abruptly end the process. This closes
    long-standing issue 310_ as well as its duplicates: issue 312, issue 1673, issue 1845, and issue 1941.

  • patch = subprocess measures coverage in Python subprocesses created
    with :mod:subprocess, :func:os.system, or one of the :func:execv or :func:spawnv family of
    functions. Closes old issue 367, its duplicate issue 378 and old
    issue 689_.

  • patch = execv adjusts the :func:execv family of
    functions to save coverage data before ending the current program and
    starting the next. Not available on Windows. Closes issue 43_ after 15
    years!

• The HTML report now dimly colors subsequent lines in multi-line statements.
  They used to have no color. This gives a better indication of the amount of
  code missing in the report. Closes issue 1308_.

• Two new exclusion patterns are part of the defaults: ... is automatically
  excluded as a line and if TYPE_CHECKING: is excluded as a branch. Closes
  issue 831_.

• A new command-line option: --save-signal=USR1 specifies a signal that
  coverage.py will listen for. When the signal is sent, the coverage data will
  be saved. This makes it possible to save data from within long-running
  processes. Thanks, Arkady Gilinsky _.

• A new configuration option: ":ref:config_report_partial_also" is a list of
  regexes to add as pragmas for partial branches. This parallels the
  ":ref:config_report_exclude_also" setting for adding line exclusion
  patterns.

• A few file path configuration settings didn't allow for tilde expansion:
  :ref:config_json_output, :ref:config_lcov_output and
  :ref:config_run_debug_file. This is now fixed.

• Wheels are included for 3.14 now that 3.14 rc1 is available.

• We no longer ship a PyPy-specific wheel. PyPy will install the pure-Python
  wheel. Closes issue 2001_.

• In the very unusual situation of not having a current frame, coverage no
  longer crashes when using the sysmon core, fixing issue 2005_.

.. _issue 43: # 43
.. _issue 310: # 310
.. _issue 312: # 312
.. _issue 367: # 367
.. _issue 378: # 378
.. _issue 689: # 689
.. _issue 831: # 831
.. _issue 1308: # 1308
.. _issue 1673: # 1673
.. _issue 1845: # 1845
.. _issue 1941: # 1941
.. _pull 1998: # 1998
.. _issue 2001: # 2001
.. _issue 2005: # 2005

.. _changes_7-9-2:

v7.9.2

Compare Source

• Fix: complex conditionals within a line might cause a KeyError when using
  sys.monitoring, as reported in issue 1991_. This is now fixed.

• Fix: we can now measure coverage for code in Python archive (.par) files.
  Thanks, Itamer Oren _.

.. _pull 1984: # 1984
.. _issue 1991: # 1991

.. _changes_7-9-1:

v7.9.1

Compare Source

• The "no-ctracer" warning is not issued for Python pre-release versions.
  Coverage doesn't ship compiled wheels for those versions, so this was far too
  noisy.

• On Python 3.14+, the "sysmon" core is now the default if it's supported for
  your configuration. Plugins and dynamic contexts are still not supported
  with it.

.. _changes_7-9-0:

v7.9.0

Compare Source

• Added a [run] core configuration setting to specify the measurement core,
  which was previously only available through the COVERAGE_CORE environment
  variable. Finishes issue 1746_.

• Fixed incorrect rendering of f-strings with doubled braces, closing issue 1980_.

• If the C tracer core can't be imported, a warning ("no-ctracer") is issued
  with the reason.

• The C tracer core extension module now conforms to PEP 489, closing issue 1977. Thanks, Adam Turner _.

• Fixed a "ValueError: min() arg is an empty sequence" error caused by strange
  empty modules, found by oss-fuzz_.

.. _issue 1746: # 1746
.. _issue 1977: # 1977
.. _pull 1978: # 1978
.. _issue 1980: # 1980
.. _PEP 489: https://peps.python.org/pep-0489
.. _oss-fuzz: https://google.github.io/oss-fuzz/

.. _changes_7-8-2:

v7.8.2

Compare Source

• Wheels are provided for Windows ARM64 on Python 3.11, 3.12, and 3.13.
  Thanks, Finn Womack _.

.. _issue 1971: # 1971
.. _pull 1972: # 1972

.. _changes_7-8-1:

v7.8.1

Compare Source

• A number of EncodingWarnings were fixed that could appear if you've enabled
  PYTHONWARNDEFAULTENCODING, fixing issue 1966. Thanks, Henry Schreiner .

• Fixed a race condition when using sys.monitoring with free-threading Python,
  closing issue 1970_.

.. _issue 1966: # 1966
.. _pull 1967: # 1967
.. _issue 1970: # 1970

.. _changes_7-8-0:

v7.8.0

Compare Source

• Added a new source_dirs setting for symmetry with the existing
  source_pkgs setting. It's preferable to the existing source setting,
  because you'll get a clear error when directories don't exist. Fixes issue 1942. Thanks, Jeremy Fleischman .

• Fix: the PYTHONSAFEPATH environment variable new in Python 3.11 is properly
  supported, closing issue 1696. Thanks, Philipp A. . This
  works properly except for a detail when using the coverage command on
  Windows. There you can use python -m coverage instead if you need exact
  emulation.

.. _issue 1696: # 1696
.. _pull 1700: # 1700
.. _issue 1942: # 1942
.. _pull 1943: # 1943

.. _changes_7-7-1:

v7.7.1

Compare Source

• A few small tweaks to the sys.monitoring support for Python 3.14. Please
  test!

.. _changes_7-7-0:

v7.7.0

Compare Source

• The Coverage object has a new method, :meth:.Coverage.branch_stats for
  getting simple branch information for a module. Closes issue 1888_.

• The :class:Coverage constructor<.Coverage> now has a plugins parameter
  for passing in plugin objects directly, thanks to Alex Gaynor _.

• Many constant tests in if statements are now recognized as being optimized
  away. For example, previously if 13: would have been considered a branch
  with one path not taken. Now it is understood as always true and no coverage
  is missing.

• The experimental sys.monitoring support now works for branch coverage if you
  are using Python 3.14.0 alpha 6 or newer. This should reduce the overhead
  coverage.py imposes on your test suite. Set the environment variable
  COVERAGE_CORE=sysmon to try it out.

• Confirmed support for PyPy 3.11. Thanks Michal Gorny.

.. _issue 1888: # 1888
.. _pull 1919: # 1919

.. _changes_7-6-12:

v7.6.12

Compare Source

• Fix: some aarch64 distributions were missing (issue 1927_). These are now
  building reliably.

.. _issue 1927: # 1927

.. _changes_7-6-11:

v7.6.11

Compare Source

• Fix: a memory leak in CTracer has been fixed. The details are in issue 1924_ and pytest-dev 676_. This should reduce the memory footprint for
  everyone even if it hadn't caused a problem before.

• We now ship a py3-none-any.whl wheel file. Thanks, Russell Keith-Magee _.

.. _pull 1914: # 1914
.. _issue 1924: # 1924
.. _pytest-dev 676: pytest-dev/pytest-cov#676

.. _changes_7-6-10:

v7.6.10

Compare Source

• Fix: some descriptions of missing branches in HTML and LCOV reports were
  incorrect when multi-line statements were involved (issue 1874_ and issue 1875_). These are now fixed.

• Fix: Python 3.14 defers evaluation of annotations _ by moving them
  into separate code objects. That code is rarely executed, so coverage.py
  would mark them as missing, as reported in issue 1908_. Now they are
  ignored by coverage automatically.

• Fixed an obscure and mysterious problem on PyPy 3.10 seemingly involving
  mocks, imports, and trace functions: issue 1902_. To be honest, I don't
  understand the problem or the solution, but git bisect helped find it,
  and now it's fixed.

• Docs: re-wrote the :ref:subprocess page to put multiprocessing first and to
  highlight the correct use of :class:multiprocessing.Pool .

.. _issue 1874: # 1874
.. _issue 1875: # 1875
.. _issue 1902: # 1902
.. _issue 1908: # 1908
.. _pep649: https://docs.python.org/3.14/whatsnew/3.14.html#whatsnew314-deferred-annotations

.. _changes_7-6-9:

v7.6.9

Compare Source

• Fix: Tomas Uribe fixed _ a performance problem in the XML
  report. Large code bases should produce XML reports much faster now.

.. _pull 1901: # 1901

.. _changes_7-6-8:

v7.6.8

Compare Source

• Fix: the LCOV report code assumed that a branch line that took no branches
  meant that the entire line was unexecuted. This isn't true in a few cases:
  the line might always raise an exception, or might have been optimized away.
  Fixes issue 1896_.

• Fix: similarly, the HTML report will now explain that a line that jumps to
  none of its expected destinations must have always raised an exception.
  Previously, it would say something nonsensical like, "line 4 didn't jump to
  line 5 because line 4 was never true, and it didn't jump to line 7 because
  line 4 was always true." This was also shown in issue 1896_.

.. _issue 1896: # 1896

.. _changes_7-6-7:

v7.6.7

Compare Source

• Fix: ugh, the other assert from 7.6.5 can also be encountered in the wild,
  so it's been restored to a conditional. Sorry for the churn.

.. _changes_7-6-6:

v7.6.6

Compare Source

• One of the new asserts from 7.6.5 caused problems in real projects, as
  reported in issue 1891_. The assert has been removed.

.. _issue 1891: # 1891

.. _changes_7-6-5:

v7.6.5

Compare Source

• Fix: fine-tuned the exact Python version (3.12.6) when exiting from with
  statements changed how they traced. This affected whether people saw the
  fix for issue 1880_.

• Fix: isolate our code more from mocking in the os module that in rare cases
  can cause bizarre behavior _.

• Refactor: some code unreachable code paths in parser.py were changed to
  asserts. If you encounter any of these, please let me know!

.. _pytest-cov-666: pytest-dev/pytest-cov#666

.. _changes_7-6-4:

v7.6.4

Compare Source

• Fix: multi-line with statements could cause contained branches to be
  incorrectly marked as missing (issue 1880_). This is now fixed.

.. _issue 1880: # 1880

.. _changes_7-6-3:

v7.6.3

Compare Source

• Fix: nested context managers could incorrectly be analyzed to flag a missing
  branch on the last context manager, as described in issue 1876_. This is
  now fixed.

• Fix: the missing branch message about not exiting a module had an extra
  "didn't," as described in issue 1873_. This is now fixed.

.. _issue 1873: # 1873
.. _issue 1876: # 1876

.. _changes_7-6-2:

v7.6.2

Compare Source

• Dropped support for Python 3.8 and PyPy 3.8.

• Fix: a final wildcard match/case clause assigning to a name (case _ as value) was incorrectly marked as a missing branch. This is now fixed,
  closing issue 1860_.

• Fewer things are considered branches now. Lambdas, comprehensions, and
  generator expressions are no longer marked as missing branches if they don't
  complete execution. Closes issue 1852_.

• Fix: the HTML report didn't properly show multi-line f-strings that end with
  a backslash continuation. This is now fixed, closing issue 1836, thanks
  to LiuYinCarl and Marco Ricci .

• Fix: the LCOV report now has correct line numbers (fixing issue 1846) and
  better branch descriptions for BRDA records (fixing issue 1850). There
  are other changes to lcov also, including a new configuration option
  :ref:line_checksums to control whether line
  checksums are included in the lcov report. The default is false. To keep
  checksums set it to true. All this work is thanks to Zack Weinberg
  (pull 1849_ and pull 1851_).

• Fixed the docs for multi-line regex exclusions, closing issue 1863_.

• Fixed a potential crash in the C tracer, closing issue 1835, thanks to
  Jan Kuhle .

.. _issue 1835: # 1835
.. _issue 1836: # 1836
.. _pull 1838: # 1838
.. _pull 1843: # 1843
.. _issue 1846: # 1846
.. _pull 1849: # 1849
.. _issue 1850: # 1850
.. _pull 1851: # 1851
.. _issue 1852: # 1852
.. _issue 1860: # 1860
.. _issue 1863: # 1863

.. _changes_7-6-1:

v7.6.1

Compare Source

• Fix: some aarch64 distributions were missing (issue 1927_). These are now
  building reliably.

.. _issue 1927: # 1927

.. _changes_7-6-11:

v7.6.0

Compare Source

• Exclusion patterns can now be multi-line, thanks to Daniel Diniz . This enables many interesting exclusion use-cases, including those
  requested in issues 118 (entire files), 996 _ (multiple lines only when appearing together), 1741 _ (remainder of a function), and 1803 _
  (arbitrary sequence of marked lines). See the :ref:multi_line_exclude
  section of the docs for more details and examples.

• The JSON report now includes per-function and per-class coverage information.
  Thanks to Daniel Diniz _ for getting the work started. This
  closes issue 1793_ and issue 1532_.

• Fixed an incorrect calculation of "(no class)" lines in the HTML classes
  report.

• Python 3.13.0b3 is supported.

.. _issue 118: # 118
.. _issue 996: # 996
.. _issue 1532: # 1532
.. _issue 1741: # 1741
.. _issue 1793: # 1793
.. _issue 1803: # 1803
.. _pull 1807: # 1807
.. _pull 1809: # 1809

.. _changes_7-5-4:

v7.5.4

Compare Source

• If you attempt to combine statement coverage data with branch coverage data,
  coverage.py used to fail with the message "Can't combine arc data with line
  data" or its reverse, "Can't combine line data with arc data." These
  messages used internal terminology, making it hard for people to understand
  the problem. They are now changed to mention "branch coverage data" and
  "statement coverage data."

• Fixed a minor branch coverage problem with wildcard match/case cases using
  names or guard clauses.

• Started testing on 3.13 free-threading (nogil) builds of Python. I'm not
  claiming full support yet. Closes issue 1799_.

.. _issue 1799: # 1799

.. _changes_7-5-3:

v7.5.3

Compare Source

• Performance improvements for combining data files, especially when measuring
  line coverage. A few different quadratic behaviors were eliminated. In one
  extreme case of combining 700+ data files, the time dropped from more than
  three hours to seven minutes. Thanks for Kraken Tech for funding the fix.

• Performance improvements for generating HTML reports, with a side benefit of
  reducing memory use, closing issue 1791_. Thanks to Daniel Diniz for
  helping to diagnose the problem.

.. _issue 1791: # 1791

.. _changes_7-5-2:

v7.5.2

Compare Source

• Fix: nested matches of exclude patterns could exclude too much code, as
  reported in issue 1779_. This is now fixed.

• Changed: previously, coverage.py would consider a module docstring to be an
  executable statement if it appeared after line 1 in the file, but not
  executable if it was the first line. Now module docstrings are never counted
  as executable statements. This can change coverage.py's count of the number
  of statements in a file, which can slightly change the coverage percentage
  reported.

• In the HTML report, the filter term and "hide covered" checkbox settings are
  remembered between viewings, thanks to Daniel Diniz _.

• Python 3.13.0b1 is supported.

• Fix: parsing error handling is improved to ensure bizarre source files are
  handled gracefully, and to unblock oss-fuzz fuzzing, thanks to Liam DeVoe . Closes issue 1787.

.. _pull 1776: # 1776
.. _issue 1779: # 1779
.. _issue 1787: # 1787
.. _pull 1788: # 1788

.. _changes_7-5-1:

v7.5.1

Compare Source

• Fix: a pragma comment on the continuation lines of a multi-line statement
  now excludes the statement and its body, the same as if the pragma is
  on the first line. This closes issue 754. The fix was contributed by
  Daniel Diniz .

• Fix: very complex source files like this one _ could
  cause a maximum recursion error when creating an HTML report. This is now
  fixed, closing issue 1774_.

• HTML report improvements:

  • Support files (JavaScript and CSS) referenced by the HTML report now have
    hashes added to their names to ensure updated files are used instead of
    stale cached copies.

  • Missing branch coverage explanations that said "the condition was never
    false" now read "the condition was always true" because it's easier to
    understand.

  • Column sort order is remembered better as you move between the index pages,
    fixing issue 1766. Thanks, Daniel Diniz .

.. _resolvent_lookup: https://github.com/sympy/sympy/blob/130950f3e6b3f97fcc17f4599ac08f70fdd2e9d4/sympy/polys/numberfields/resolvent_lookup.py
.. _issue 754: # 754
.. _issue 1766: # 1766
.. _pull 1768: # 1768
.. _pull 1773: # 1773
.. _issue 1774: # 1774

.. _changes_7-5-0:

v7.5.0

Compare Source

• Added initial support for function and class reporting in the HTML report.
  There are now three index pages which link to each other: files, functions,
  and classes. Other reports don't yet have this information, but it will be
  added in the future where it makes sense. Feedback gladly accepted!
  Finishes issue 780_.

• Other HTML report improvements:

  • There is now a "hide covered" checkbox to filter out 100% files, finishing
    issue 1384_.

  • The index page is always sorted by one of its columns, with clearer
    indications of the sorting.

  • The "previous file" shortcut key didn't work on the index page, but now it
    does, fixing issue 1765_.

• The debug output showing which configuration files were tried now shows
  absolute paths to help diagnose problems where settings aren't taking effect,
  and is renamed from "attempted_config_files" to the more logical
  "config_files_attempted."

• Python 3.13.0a6 is supported.

.. _issue 780: # 780
.. _issue 1384: # 1384
.. _issue 1765: # 1765

.. _changes_7-4-4:

v7.4.4

Compare Source

• Fix: in some cases, even with [run] relative_files=True, a data file
  could be created with absolute path names. When combined with other relative
  data files, it was random whether the absolute file names would be made
  relative or not. If they weren't, then a file would be listed twice in
  reports, as detailed in issue 1752_. This is now fixed: absolute file
  names are always made relative when combining. Thanks to Bruno Rodrigues dos
  Santos for support.

• Fix: the last case of a match/case statement had an incorrect message if the
  branch was missed. It said the pattern never matched, when actually the
  branch is missed if the last case always matched.

• Fix: clicking a line number in the HTML report now positions more accurately.

• Fix: the report:format setting was defined as a boolean, but should be a
  string. Thanks, Tanaydin Sirin _. It is also now documented
  on the :ref:configuration page .

.. _issue 1752: # 1752
.. _pull 1754: # 1754

.. _changes_7-4-3:

v7.4.3

Compare Source

• Fix: in some cases, coverage could fail with a RuntimeError: "Set changed
  size during iteration." This is now fixed, closing issue 1733_.

.. _issue 1733: # 1733

.. _changes_7-4-2:

v7.4.2

Compare Source

• Fix: setting COVERAGE_CORE=sysmon no longer errors on 3.11 and lower,
  thanks Hugo van Kemenade _. It now issues a warning that
  sys.monitoring is not available and falls back to the default core instead.

.. _pull 1747: # 1747

.. _changes_7-4-1:

v7.4.1

Compare Source

• Python 3.13.0a3 is supported.

• Fix: the JSON report now includes an explicit format version number, closing
  issue 1732_.

.. _issue 1732: # 1732

.. _changes_7-4-0:

v7.4.0

Compare Source

• In Python 3.12 and above, you can try an experimental core based on the new
  :mod:sys.monitoring module by defining a
  COVERAGE_CORE=sysmon environment variable. This should be faster for
  line coverage, but not for branch coverage, and plugins and dynamic contexts
  are not yet supported with it. I am very interested to hear how it works (or
  doesn't!) for you.

.. _changes_7-3-4:

v7.3.4

Compare Source

• Fix: the change for multi-line signature exclusions in 7.3.3 broke other
  forms of nested clauses being excluded properly. This is now fixed, closing
  issue 1713_.

• Fix: in the HTML report, selecting code for copying won't select the line
  numbers also. Thanks, Robert Harris _.

.. _issue 1713: # 1713
.. _pull 1717: # 1717

.. _changes_7-3-3:

v7.3.3

Compare Source

• Fix: function definitions with multi-line signatures can now be excluded by
  matching any of the lines, closing issue 684. Thanks, Jan Rusak, Maciej Kowalczyk and Joanna Ejzel .

• Fix: XML reports could fail with a TypeError if files had numeric components
  that were duplicates except for leading zeroes, like file1.py and
  file001.py. Fixes issue 1709_.

• The coverage annotate command used to announce that it would be removed
  in a future version. Enough people got in touch to say that they use it, so
  it will stay. Don't expect it to keep up with other new features though.

• Added new :ref:debug options :

  • pytest writes the pytest test name into the debug output.

  • dataop2 writes the full data being added to CoverageData objects.

.. _issue 684: # 684
.. _pull 1705: # 1705
.. _issue 1709: # 1709

.. _changes_7-3-2:

v7.3.2

Compare Source

• The coverage lcov command ignored the [report] exclude_lines and
  [report] exclude_also settings (issue 1684). This is now fixed,
  thanks Jacqueline Lee .

• Sometimes SQLite will create journal files alongside the coverage.py database
  files. These are ephemeral, but could be mistakenly included when combining
  data files. Now they are always ignored, fixing issue 1605_. Thanks to
  Brad Smith for suggesting fixes and providing detailed debugging.

• On Python 3.12+, we now disable SQLite writing journal files, which should be
  a little faster.

• The new 3.12 soft keyword type is properly bolded in HTML reports.

• Removed the "fullcoverage" feature used by CPython to measure the coverage of
  early-imported standard library modules. CPython stopped using it <88054_>_ in 2021, and it stopped working completely in Python 3.13.

.. _issue 1605: # 1605
.. _issue 1684: # 1684
.. _pull 1685: # 1685
.. _88054: python/cpython#88054

.. _changes_7-3-1:

v7.3.1

Compare Source

• The semantics of stars in file patterns has been clarified in the docs. A
  leading or trailing star matches any number of path components, like a double
  star would. This is different than the behavior of a star in the middle of a
  pattern. This discrepancy was identified by Sviatoslav Sydorenko , who provided patient detailed diagnosis and
  graciously agreed to a pragmatic resolution.

• The API docs were missing from the last version. They are now restored _.

.. _apidocs: https://coverage.readthedocs.io/en/latest/api_coverage.html
.. _starbad: # 1407 (comment)
.. _pull 1650: # 1650

.. _changes_7-3-0:

v7.3.0

Compare Source

• Added a :meth:.Coverage.collect context manager to start and stop coverage
  data collection.

• Dropped support for Python 3.7.

• Fix: in unusual circumstances, SQLite cannot be set to asynchronous mode.
  Coverage.py would fail with the error Safety level may not be changed inside a transaction. This is now avoided, closing issue 1646_. Thanks
  to Michael Bell for the detailed bug report.

• Docs: examples of configuration files now include separate examples for the
  different syntaxes: .coveragerc, pyproject.toml, setup.cfg, and tox.ini.

• Fix: added nosemgrep comments to our JavaScript code so that
  semgrep-based SAST security checks won't raise false alarms about security
  problems that aren't problems.

• Added a CITATION.cff file, thanks to Ken Schackart _.

.. _pull 1641: # 1641
.. _issue 1646: # 1646

.. _changes_7-2-7:

v7.2.7

Compare Source

• Fix: reverted a change from 6.4.3 _ that helped Cython, but
  also increased the size of data files when using dynamic contexts, as
  described in the now-fixed issue 1586. The problem is now avoided due to a
  recent change (issue 1538 ). Thanks to Anders Kaseorg _ and David Szotten for persisting with problem reports and
  detailed diagnoses.

• Wheels are now provided for CPython 3.12.

.. _pull 1347b: [# 1347](https:/