Repository navigation
New distribution [0.14.2] - #28
Merged
Merged
Conversation
JarryShaw
commented
Mar 28, 2019
JarryShaw
commented
Owner
- bugfix in exceptions & TCP reassembly algo.
- moved venv in project
* bugfix in exceptions & TCP reassembly algo. * moved venv in project
JarryShaw
added
ci
JarryShaw
added a commit
that referenced
this pull request
Sep 30, 2026
Three settled rulings fit none of the four code-convention pages, because they govern the repository rather than the library. The owner ruled on #918 that they get a fifth page rather than staying in their threads. - Add ``docs/source/contributing/conventions/process.rst`` (``.. _process:``), covering what the ``all`` extra carries (#910), what a changelog entry is, and what the issue and pull request labels mean. Paraphrased throughout rather than quoting the owner, on his instruction on the same issue. - Record the changelog grouping as ruled: a section per top-level module with ``Added``/``Changed``/``Fixed`` nested inside each, at module granularity. ``docs/source/changelog/1.5.0.rst`` is currently the inverse -- three flat kind-runs -- and the restructure belongs to #657, which owns the file and merges last. The one case the rule does not settle, an entry spanning modules, is flagged rather than decided. - Separate the two axes the changelog ruling is easy to conflate with: a pull request's commits, where #657 deliberately carries roughly one per code pull request, versus a changelog file's entries, which group. The page gives the commands for both rather than figures that go stale. - Document the label scheme the owner asked for alongside ``breaking``: the type labels matching the commit subject prefix, the issue-kind labels, the state labels, the head-keyed ``review:`` family, and dependabot's scope labels. It says plainly that labels are applied by hand -- nothing automates them except dependabot's own -- because that is the part readers assume wrongly. - Correct two claims about ``breaking`` that measurement contradicts. It **is** applied below #350: the carriers are ``#3``-``#28``, seven pre-``0.15`` distribution pull requests, with nothing between ``#28`` and ``#350``; on issues it starts at ``#775``. And no ruling defines the label's meaning, so the page states the behavioural test -- whether a caller can observe the difference without changing their code -- rather than attributing a phrasing to anyone. - Add ``process`` to ``ANCHORS`` and the index toctree. A fifth page breaks every four-counting assertion in ``tests/project/test_conventions_doc_claims.py``; those are desensitised, and ``ProcessConventionTests`` adds seven tests pinning the page's claims against the tree -- the extras in ``pyproject.toml``, the changelog file's real shape, the module packages it names, and the template's commit-type tickboxes. tests/project: 217 passed, 1 skipped, 604 subtests (210/572 before). Every new assertion was shown to fail with its claim removed or contradicted.
JarryShaw
added a commit
that referenced
this pull request
Sep 30, 2026
Three settled rulings fit none of the four code-convention pages, because they govern the repository rather than the library. The owner ruled on #918 that they get a fifth page rather than staying in their threads. - Add ``docs/source/contributing/conventions/process.rst`` (``.. _process:``), covering what the ``all`` extra carries (#910), what a changelog entry is, and what the issue and pull request labels mean. Paraphrased throughout rather than quoting the owner, on his instruction on the same issue. - Record the changelog grouping as ruled: a section per top-level module with ``Added``/``Changed``/``Fixed`` nested inside each, at module granularity. The restructure belongs to #657, which owns the file and merges last. The one case the rule does not settle, an entry spanning modules, is flagged rather than decided. - Describe the file's **current** shape from measurement. An earlier draft called it "three flat kind-runs", carried from a comment nobody had checked; the 80 entries actually carry the three kind labels in **26** runs, blocked for roughly the first 40% of the file and interleaved after it. That is the difference between transposing three blocks and regrouping scattered entries, which is the work #657 has to do. - Separate the two axes the changelog ruling is easy to conflate with: a pull request's commits, where #657 deliberately carries roughly one per code pull request, versus a changelog file's entries, which group. The page gives the commands for both rather than figures that go stale. - Document the label scheme the owner asked for alongside ``breaking``: the type labels matching the commit subject prefix, the issue-kind labels, the state labels, the head-keyed ``review:`` family, and dependabot's scope labels. - Name **both** automated label paths, not one. An earlier draft said dependabot was the single exception to hand-application; the issue templates are a second, applying a label from front matter before anyone reads the issue -- ``.github/ISSUE_TEMPLATE/bug_report.md`` carries ``labels: bug`` and ``feature_request.md`` carries ``labels: enhancement``. So ``bug`` and ``enhancement`` on a template-opened issue are defaults rather than assessments. ``.github/release.yml`` only reads labels, and ``PCAPKIT_CONDA_LABEL`` is a conda channel label; both ruled out explicitly. - Correct two claims about ``breaking`` that measurement contradicts. It **is** applied below #350: the carriers are ``#3``-``#28``, seven pre-``0.15`` distribution pull requests, with nothing between ``#28`` and ``#350``; on issues it starts at ``#775``. And no ruling defines the label's meaning, so the page states the behavioural test -- whether a caller can observe the difference without changing their code -- rather than attributing a phrasing to anyone. - Add ``process`` to ``ANCHORS`` and the index toctree. A fifth page breaks every four-counting assertion in ``tests/project/test_conventions_doc_claims.py``; those are desensitised, and ``ProcessConventionTests`` adds eight tests pinning the page's claims against the tree. Two of them exist because the first versions could not see the defects above: the label test now sweeps ``.github/ISSUE_TEMPLATE/*.md`` for front-matter labels rather than reading ``dependabot.yml`` alone, and the new run-structure test asserts the measured count *in the page's prose* rather than the bare figure, which the page's own ``wc -l`` output had satisfied. tests/project: 218 passed, 1 skipped, 606 subtests (210/572 before). Every new assertion was shown to fail with its claim removed or contradicted.
JarryShaw
added a commit
that referenced
this pull request
Sep 30, 2026
Three settled rulings fit none of the four code-convention pages, because they govern the repository rather than the library. The owner ruled on #918 that they get a fifth page rather than staying in their threads. - Add ``docs/source/contributing/conventions/process.rst`` (``.. _process:``), covering what the ``all`` extra carries (#910), what a changelog entry is, and what the issue and pull request labels mean. Paraphrased throughout rather than quoting the owner, on his instruction on the same issue. - Record the changelog grouping as ruled: a section per top-level module with ``Added``/``Changed``/``Fixed`` nested inside each, at module granularity. The restructure belongs to #657, which owns the file and merges last. The one case the rule does not settle, an entry spanning modules, is flagged rather than decided. - Describe the file's **current** shape from measurement. An earlier draft called it "three flat kind-runs", carried from a comment nobody had checked; the 80 entries actually carry the three kind labels in **26** runs, blocked for roughly the first 40% of the file and interleaved after it. That is the difference between transposing three blocks and regrouping scattered entries, which is the work #657 has to do. - Separate the two axes the changelog ruling is easy to conflate with: a pull request's commits, where #657 deliberately carries roughly one per code pull request, versus a changelog file's entries, which group. The page gives the commands for both rather than figures that go stale. - Document the label scheme the owner asked for alongside ``breaking``: the type labels matching the commit subject prefix, the issue-kind labels, the state labels, the head-keyed ``review:`` family, and dependabot's scope labels. - Name **both** automated label paths, not one. An earlier draft said dependabot was the single exception to hand-application; the issue templates are a second, applying a label from front matter before anyone reads the issue -- ``.github/ISSUE_TEMPLATE/bug_report.md`` carries ``labels: bug`` and ``feature_request.md`` carries ``labels: enhancement``. So ``bug`` and ``enhancement`` on a template-opened issue are defaults rather than assessments. ``.github/release.yml`` only reads labels, and ``PCAPKIT_CONDA_LABEL`` is a conda channel label; both ruled out explicitly. - Correct two claims about ``breaking`` that measurement contradicts. It **is** applied below #350: the carriers are ``#3``-``#28``, seven pre-``0.15`` distribution pull requests, with nothing between ``#28`` and ``#350``; on issues it starts at ``#775``. And no ruling defines the label's meaning, so the page states the behavioural test -- whether a caller can observe the difference without changing their code -- rather than attributing a phrasing to anyone. - Add ``process`` to ``ANCHORS`` and the index toctree. A fifth page breaks every four-counting assertion in ``tests/project/test_conventions_doc_claims.py``; those are desensitised, and ``ProcessConventionTests`` adds eight tests pinning the page's claims against the tree. Two of them exist because the first versions could not see the defects above: the label test now sweeps ``.github/ISSUE_TEMPLATE/*.md`` for front-matter labels rather than reading ``dependabot.yml`` alone, and the new run-structure test asserts the measured count *in the page's prose* rather than the bare figure, which the page's own ``wc -l`` output had satisfied. tests/project: 218 passed, 1 skipped, 606 subtests (210/572 before). Every new assertion was shown to fail with its claim removed or contradicted.
JarryShaw
added a commit
that referenced
this pull request
Sep 30, 2026
Three settled rulings fit none of the four code-convention pages, because they govern the repository rather than the library. The owner ruled on #918 that they get a fifth page rather than staying in their threads. - Add ``docs/source/contributing/conventions/process.rst`` (``.. _process:``), covering what the ``all`` extra carries (#910), what a changelog entry is, and what the issue and pull request labels mean. Paraphrased throughout rather than quoting the owner, on his instruction on the same issue. - Record the changelog grouping as ruled: a section per top-level module with ``Added``/``Changed``/``Fixed`` nested inside each, at module granularity. The restructure belongs to #657, which owns the file and merges last. The one case the rule does not settle, an entry spanning modules, is flagged rather than decided. - Describe the file's **current** shape from measurement. An earlier draft called it "three flat kind-runs", carried from a comment nobody had checked; the 80 entries actually carry the three kind labels in **26** runs, blocked for roughly the first 40% of the file and interleaved after it. That is the difference between transposing three blocks and regrouping scattered entries, which is the work #657 has to do. - Separate the two axes the changelog ruling is easy to conflate with: a pull request's commits, where #657 deliberately carries roughly one per code pull request, versus a changelog file's entries, which group. The page gives the commands for both rather than figures that go stale. - Document the label scheme the owner asked for alongside ``breaking``: the type labels matching the commit subject prefix, the issue-kind labels, the state labels, the head-keyed ``review:`` family, and dependabot's scope labels. - Name **both** automated label paths, not one. An earlier draft said dependabot was the single exception to hand-application; the issue templates are a second, applying a label from front matter before anyone reads the issue -- ``.github/ISSUE_TEMPLATE/bug_report.md`` carries ``labels: bug`` and ``feature_request.md`` carries ``labels: enhancement``. So ``bug`` and ``enhancement`` on a template-opened issue are defaults rather than assessments. ``.github/release.yml`` only reads labels, and ``PCAPKIT_CONDA_LABEL`` is a conda channel label; both ruled out explicitly. - Correct two claims about ``breaking`` that measurement contradicts. It **is** applied below #350: the carriers are ``#3``-``#28``, seven pre-``0.15`` distribution pull requests, with nothing between ``#28`` and ``#350``; on issues it starts at ``#775``. And no ruling defines the label's meaning, so the page states the behavioural test -- whether a caller can observe the difference without changing their code -- rather than attributing a phrasing to anyone. - Add ``process`` to ``ANCHORS`` and the index toctree. A fifth page breaks every four-counting assertion in ``tests/project/test_conventions_doc_claims.py``; those are desensitised, and ``ProcessConventionTests`` adds eight tests pinning the page's claims against the tree. Two of them exist because the first versions could not see the defects above: the label test now sweeps ``.github/ISSUE_TEMPLATE/*.md`` for front-matter labels rather than reading ``dependabot.yml`` alone, and the new run-structure test asserts the measured count *in the page's prose* rather than the bare figure, which the page's own ``wc -l`` output had satisfied. tests/project: 218 passed, 1 skipped, 606 subtests (210/572 before). Every new assertion was shown to fail with its claim removed or contradicted.
JarryShaw
added a commit
that referenced
this pull request
Sep 30, 2026
Three settled rulings fit none of the four code-convention pages, because they govern the repository rather than the library. The owner ruled on #918 that they get a fifth page rather than staying in their threads. - Add ``docs/source/contributing/conventions/process.rst`` (``.. _process:``), covering what the ``all`` extra carries (#910), what a changelog entry is, and what the issue and pull request labels mean. Paraphrased throughout rather than quoting the owner, on his instruction on the same issue. - Record the changelog grouping as ruled: a section per top-level module with ``Added``/``Changed``/``Fixed`` nested inside each, at module granularity. The restructure belongs to #657, which owns the file and merges last. The one case the rule does not settle, an entry spanning modules, is flagged rather than decided. - Describe the file's **current** shape from measurement. An earlier draft called it "three flat kind-runs", carried from a comment nobody had checked; the 80 entries actually carry the three kind labels in **26** runs, blocked for roughly the first 40% of the file and interleaved after it. That is the difference between transposing three blocks and regrouping scattered entries, which is the work #657 has to do. - Separate the two axes the changelog ruling is easy to conflate with: a pull request's commits, where #657 deliberately carries roughly one per code pull request, versus a changelog file's entries, which group. The page gives the commands for both rather than figures that go stale. - Document the label scheme the owner asked for alongside ``breaking``: the type labels matching the commit subject prefix, the issue-kind labels, the state labels, the head-keyed ``review:`` family, and dependabot's scope labels. - Name **both** automated label paths, not one. An earlier draft said dependabot was the single exception to hand-application; the issue templates are a second, applying a label from front matter before anyone reads the issue -- ``.github/ISSUE_TEMPLATE/bug_report.md`` carries ``labels: bug`` and ``feature_request.md`` carries ``labels: enhancement``. So ``bug`` and ``enhancement`` on a template-opened issue are defaults rather than assessments. ``.github/release.yml`` only reads labels, and ``PCAPKIT_CONDA_LABEL`` is a conda channel label; both ruled out explicitly. - Correct two claims about ``breaking`` that measurement contradicts. It **is** applied below #350: the carriers are ``#3``-``#28``, seven pre-``0.15`` distribution pull requests, with nothing between ``#28`` and ``#350``; on issues it starts at ``#775``. And no ruling defines the label's meaning, so the page states the behavioural test -- whether a caller can observe the difference without changing their code -- rather than attributing a phrasing to anyone. - Add ``process`` to ``ANCHORS`` and the index toctree. A fifth page breaks every four-counting assertion in ``tests/project/test_conventions_doc_claims.py``; those are desensitised, and ``ProcessConventionTests`` adds eight tests pinning the page's claims against the tree. Two of them exist because the first versions could not see the defects above: the label test now sweeps ``.github/ISSUE_TEMPLATE/*.md`` for front-matter labels rather than reading ``dependabot.yml`` alone, and the new run-structure test asserts the measured count *in the page's prose* rather than the bare figure, which the page's own ``wc -l`` output had satisfied. tests/project: 218 passed, 1 skipped, 606 subtests (210/572 before). Every new assertion was shown to fail with its claim removed or contradicted.
JarryShaw
added a commit
that referenced
this pull request
Sep 30, 2026
Three settled rulings fit none of the four code-convention pages, because they govern the repository rather than the library. The owner ruled on #918 that they get a fifth page rather than staying in their threads. - Add ``docs/source/contributing/conventions/process.rst`` (``.. _process:``), covering what the ``all`` extra carries (#910), what a changelog entry is, and what the issue and pull request labels mean. Paraphrased throughout rather than quoting the owner, on his instruction on the same issue. - Record the changelog grouping as ruled: a section per top-level module with ``Added``/``Changed``/``Fixed`` nested inside each. The restructure belongs to #657, which owns the file and merges last. The one case the rule does not settle, an entry spanning modules, is flagged rather than decided. - Describe the file's **current** shape from measurement. An earlier draft called it "three flat kind-runs", carried from a comment nobody had checked; the 80 entries actually carry the three kind labels in **26** runs. - Document the label scheme the owner asked for alongside ``breaking``, and name **both** automated paths: dependabot, and the issue templates, which apply a label from front matter before anyone reads the issue. An earlier draft called dependabot the single exception. - Correct two ``breaking`` claims measurement contradicts. It **is** applied below #350 -- carriers ``#3``-``#28``, nothing between ``#28`` and ``#350``, issues from ``#775``. And no ruling defines the label, so the page states the behavioural test rather than attributing a phrasing to anyone. - Correct the audit table's population to **six** -- the ``mh``/``ngap`` helpers ``EnumLookup`` subclasses those modules define, ``ProcedureCode``/``ProtocolIE`` being re-exports. The page had carried three figures for one population. Tests, and the reason each exists rather than the shape of it: - ``ProcessConventionTests`` pins the page's claims against the tree. Two were rewritten because they could pass for the wrong reason: one read only ``dependabot.yml`` and never ``ISSUE_TEMPLATE/``, and one asserted the bare string ``26``, which the page's own ``wc -l`` output satisfied. - ``test_a_non_string_key_raises_a_value_miss_on_every_helper`` **derives** the population instead of listing it. Its predecessor was named ``..._on_all_seven`` and exercised three. - ``test_every_issue_link_number_matches_its_own_url`` closes the largest class of unpinned claim: a cross-review corrupted ~30 link numbers and not one was caught, because nothing compared the displayed ``#NNN`` against the number in its own URL. Invisible to a reader, unwarnable by Sphinx. - ``_optional_dependencies`` replaces ``tomllib``, which is 3.11+ and broke the ``Python 3.10`` leg while passing locally on 3.14 -- ``tomli`` is not a dependency either, so there was nothing to fall back to. It now strips comments and tracks bracket **depth**: a version matching to the first ``]`` mis-parsed **3 of the 14** extras silently, emptying ``vendor`` on the ``]`` inside ``"requests[socks]"``, truncating ``dev``, and **inventing a requirement called** ``9 skipped`` **out of comment prose**. No shipped assertion read those keys, but ``dev`` and ``vendor`` are both described in the page's prose, so the next assertion to check either would have got wrong data with nothing raised. ``test_the_extras_reader_agrees_with_tomllib`` now compares the two key for key wherever a real parser exists, and skips on 3.10 rather than pretending to. - The page states plainly what cannot be pinned here: any claim whose ground truth is a GitHub query, since CI has no network. The command beside the figure is the pin, run by a reader rather than by CI. tests/project: 221 passed, 1 skipped, 640 subtests. The extras reader verified byte-identical under python3.10 and 3.14, and every new assertion shown to fail with its claim removed or its defect reintroduced.
JarryShaw
added a commit
that referenced
this pull request
Sep 30, 2026
Three settled rulings fit none of the four code-convention pages, because they govern the repository rather than the library. The owner ruled on #918 that they get a fifth page rather than staying in their threads. - Add ``docs/source/contributing/conventions/process.rst`` (``.. _process:``), covering what the ``all`` extra carries (#910), what a changelog entry is, and what the issue and pull request labels mean. Paraphrased throughout rather than quoting the owner, on his instruction on the same issue. - Record the changelog grouping as ruled: a section per top-level module with ``Added``/``Changed``/``Fixed`` nested inside each. The restructure belongs to #657, which owns the file and merges last. The one case the rule does not settle, an entry spanning modules, is flagged rather than decided. - Describe the file's **current** shape from measurement. An earlier draft called it "three flat kind-runs", carried from a comment nobody had checked; the 80 entries actually carry the three kind labels in **26** runs. - Document the label scheme the owner asked for alongside ``breaking``, and name **both** automated paths: dependabot, and the issue templates, which apply a label from front matter before anyone reads the issue. An earlier draft called dependabot the single exception. - Correct two ``breaking`` claims measurement contradicts. It **is** applied below #350 -- carriers ``#3``-``#28``, nothing between ``#28`` and ``#350``, issues from ``#775``. And no ruling defines the label, so the page states the behavioural test rather than attributing a phrasing to anyone. - Correct the audit table's population to **six** -- the ``mh``/``ngap`` helpers ``EnumLookup`` subclasses those modules define, ``ProcedureCode``/``ProtocolIE`` being re-exports. The page had carried three figures for one population. Tests, and the reason each exists rather than the shape of it: - ``ProcessConventionTests`` pins the page's claims against the tree. Two were rewritten because they could pass for the wrong reason: one read only ``dependabot.yml`` and never ``ISSUE_TEMPLATE/``, and one asserted the bare string ``26``, which the page's own ``wc -l`` output satisfied. - ``test_a_non_string_key_raises_a_value_miss_on_every_helper`` **derives** the population instead of listing it. Its predecessor was named ``..._on_all_seven`` and exercised three. - ``test_every_issue_link_number_matches_its_own_url`` closes the largest class of unpinned claim: a cross-review corrupted ~30 link numbers and not one was caught, because nothing compared the displayed ``#NNN`` against the number in its own URL. Invisible to a reader, unwarnable by Sphinx. - ``_optional_dependencies`` replaces ``tomllib``, which is 3.11+ and broke the ``Python 3.10`` leg while passing locally on 3.14 -- ``tomli`` is not a dependency either, so there was nothing to fall back to. It now strips comments and tracks bracket **depth**: a version matching to the first ``]`` mis-parsed **3 of the 14** extras silently, emptying ``vendor`` on the ``]`` inside ``"requests[socks]"``, truncating ``dev``, and **inventing a requirement called** ``9 skipped`` **out of comment prose**. No shipped assertion read those keys, but ``dev`` and ``vendor`` are both described in the page's prose, so the next assertion to check either would have got wrong data with nothing raised. ``test_the_extras_reader_agrees_with_tomllib`` now compares the two key for key wherever a real parser exists, and skips on 3.10 rather than pretending to. - The page states plainly what cannot be pinned here: any claim whose ground truth is a GitHub query, since CI has no network. The command beside the figure is the pin, run by a reader rather than by CI. tests/project: 221 passed, 1 skipped, 640 subtests. The extras reader verified byte-identical under python3.10 and 3.14, and every new assertion shown to fail with its claim removed or its defect reintroduced.
JarryShaw
added a commit
that referenced
this pull request
Sep 30, 2026
Three settled rulings fit none of the four code-convention pages, because they govern the repository rather than the library. The owner ruled on #918 that they get a fifth page rather than staying in their threads. - Add ``docs/source/contributing/conventions/process.rst`` (``.. _process:``), covering what the ``all`` extra carries (#910), what a changelog entry is, and what the issue and pull request labels mean. Paraphrased throughout rather than quoting the owner, on his instruction on the same issue. - Record the changelog grouping as ruled: a section per top-level module with ``Added``/``Changed``/``Fixed`` nested inside each. The restructure belongs to #657, which owns the file and merges last. The one case the rule does not settle, an entry spanning modules, is flagged rather than decided. - Describe the file's **current** shape from measurement. An earlier draft called it "three flat kind-runs", carried from a comment nobody had checked; the 80 entries actually carry the three kind labels in **26** runs. - Document the label scheme the owner asked for alongside ``breaking``, and name **both** automated paths: dependabot, and the issue templates, which apply a label from front matter before anyone reads the issue. An earlier draft called dependabot the single exception. - Correct two ``breaking`` claims measurement contradicts. It **is** applied below #350 -- carriers ``#3``-``#28``, nothing between ``#28`` and ``#350``, issues from ``#775``. And no ruling defines the label, so the page states the behavioural test rather than attributing a phrasing to anyone. - Correct the audit table's population to **six** -- the ``mh``/``ngap`` helpers ``EnumLookup`` subclasses those modules define, ``ProcedureCode``/``ProtocolIE`` being re-exports. The page had carried three figures for one population. Tests, and the reason each exists rather than the shape of it: - ``ProcessConventionTests`` pins the page's claims against the tree. Two were rewritten because they could pass for the wrong reason: one read only ``dependabot.yml`` and never ``ISSUE_TEMPLATE/``, and one asserted the bare string ``26``, which the page's own ``wc -l`` output satisfied. - ``test_a_non_string_key_raises_a_value_miss_on_every_helper`` **derives** the population instead of listing it. Its predecessor was named ``..._on_all_seven`` and exercised three. - ``test_every_issue_link_number_matches_its_own_url`` closes the largest class of unpinned claim: a cross-review corrupted ~30 link numbers and not one was caught, because nothing compared the displayed ``#NNN`` against the number in its own URL. Invisible to a reader, unwarnable by Sphinx. - ``_optional_dependencies`` replaces ``tomllib``, which is 3.11+ and broke the ``Python 3.10`` leg while passing locally on 3.14 -- ``tomli`` is not a dependency either, so there was nothing to fall back to. It now strips comments and tracks bracket **depth**: a version matching to the first ``]`` mis-parsed **3 of the 14** extras silently, emptying ``vendor`` on the ``]`` inside ``"requests[socks]"``, truncating ``dev``, and **inventing a requirement called** ``9 skipped`` **out of comment prose**. No shipped assertion read those keys, but ``dev`` and ``vendor`` are both described in the page's prose, so the next assertion to check either would have got wrong data with nothing raised. ``test_the_extras_reader_agrees_with_tomllib`` now compares the two key for key wherever a real parser exists, and skips on 3.10 rather than pretending to. - The page states plainly what cannot be pinned here: any claim whose ground truth is a GitHub query, since CI has no network. The command beside the figure is the pin, run by a reader rather than by CI. tests/project: 221 passed, 1 skipped, 640 subtests. The extras reader verified byte-identical under python3.10 and 3.14, and every new assertion shown to fail with its claim removed or its defect reintroduced.
JarryShaw
added a commit
that referenced
this pull request
Sep 30, 2026
Three settled rulings fit none of the four code-convention pages, because they govern the repository rather than the library. The owner ruled on #918 that they get a fifth page rather than staying in their threads. - Add ``docs/source/contributing/conventions/process.rst`` (``.. _process:``), covering what the ``all`` extra carries (#910), what a changelog entry is, and what the issue and pull request labels mean. Paraphrased throughout rather than quoting the owner, on his instruction on the same issue. - Record the changelog grouping as ruled: a section per top-level module with ``Added``/``Changed``/``Fixed`` nested inside each. The restructure belongs to #657, which owns the file and merges last. The one case the rule does not settle, an entry spanning modules, is flagged rather than decided. - Describe the file's **current** shape from measurement. An earlier draft called it "three flat kind-runs", carried from a comment nobody had checked; the 80 entries actually carry the three kind labels in **26** runs. - Document the label scheme the owner asked for alongside ``breaking``, and name **both** automated paths: dependabot, and the issue templates, which apply a label from front matter before anyone reads the issue. An earlier draft called dependabot the single exception. - Correct two ``breaking`` claims measurement contradicts. It **is** applied below #350 -- carriers ``#3``-``#28``, nothing between ``#28`` and ``#350``, issues from ``#775``. And no ruling defines the label, so the page states the behavioural test rather than attributing a phrasing to anyone. - Correct the audit table's population to **six** -- the ``mh``/``ngap`` helpers ``EnumLookup`` subclasses those modules define, ``ProcedureCode``/``ProtocolIE`` being re-exports. The page had carried three figures for one population. Tests, and the reason each exists rather than the shape of it: - ``ProcessConventionTests`` pins the page's claims against the tree. Two were rewritten because they could pass for the wrong reason: one read only ``dependabot.yml`` and never ``ISSUE_TEMPLATE/``, and one asserted the bare string ``26``, which the page's own ``wc -l`` output satisfied. - ``test_a_non_string_key_raises_a_value_miss_on_every_helper`` **derives** the population instead of listing it. Its predecessor was named ``..._on_all_seven`` and exercised three. - ``test_every_issue_link_number_matches_its_own_url`` closes the largest class of unpinned claim: a cross-review corrupted ~30 link numbers and not one was caught, because nothing compared the displayed ``#NNN`` against the number in its own URL. Invisible to a reader, unwarnable by Sphinx. - ``_optional_dependencies`` replaces ``tomllib``, which is 3.11+ and broke the ``Python 3.10`` leg while passing locally on 3.14 -- ``tomli`` is not a dependency either, so there was nothing to fall back to. It now strips comments and tracks bracket **depth**: a version matching to the first ``]`` mis-parsed **3 of the 14** extras silently, emptying ``vendor`` on the ``]`` inside ``"requests[socks]"``, truncating ``dev``, and **inventing a requirement called** ``9 skipped`` **out of comment prose**. No shipped assertion read those keys, but ``dev`` and ``vendor`` are both described in the page's prose, so the next assertion to check either would have got wrong data with nothing raised. ``test_the_extras_reader_agrees_with_tomllib`` now compares the two key for key wherever a real parser exists, and skips on 3.10 rather than pretending to. - The page states plainly what cannot be pinned here: any claim whose ground truth is a GitHub query, since CI has no network. The command beside the figure is the pin, run by a reader rather than by CI. tests/project: 221 passed, 1 skipped, 640 subtests. The extras reader verified byte-identical under python3.10 and 3.14, and every new assertion shown to fail with its claim removed or its defect reintroduced.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.