Skip to content

#422 - Tests: behaviors shown untested by the surviving mutants - #423

Merged
njoubert-cleverage merged 1 commit into
mainfrom
422
Nov 3, 2026
Merged

njoubert-cleverage merged 1 commit into
mainfrom
422

Conversation

@njoubert-cleverage

Copy link
Copy Markdown
Member

Fixes #422

Tests only, no change of the code. Infection run again on the files of the ticket (ProcessManager, ProcessConfiguration, xpath_evaluator, XmlFile, XmlStreamReaderTask, InputCsvReaderTask, InputFolderBrowserTask, LineReaderTask, cached): covered code MSI from 69 % to 82 % (escaped mutants from 189 to 114 of 640).

  • ProcessManagerFlowTest (new), with recording tasks: skip strategy (next tasks not executed for the item, default error output = input, error output set by the task), stop propagated through the parent tasks and from an error branch, previous state and its exception given to an error branch, context given to the tasks, flush after an iterable task (next tasks and error outputs, stopped at a blocking task), stop during the flush, empty iterable task, blocking task with and without input, process executed again after a failure.
  • ProcessConfigurationTest (new): dependency groups topologically sorted on 30 random acyclic graphs (fixed seeds), main task group (entry point, else end point, else first task), cycle not sorted.
  • xpath_evaluator: single_result / ignore_missing, too many results, subquery options, option types, libxml state restored (previous errors not reported, none left, libxml_use_internal_errors() restored).
  • XmlFile, XmlStreamReaderTask: libxml state, warnings tolerated, exact (trimmed) messages, read twice, root element, directory rejected.
  • InputCsvReaderTask base_path, InputFolderBrowserTask base_folder_path (options validated on initialization, another folder rejected while browsing), cached ttl (normalized, expiry, transformers applied once on a miss, save failure logged, non-string key), LineReaderTask filename type.

Remaining mutants are mostly equivalent: visibility, log messages and contexts, the READ_AHEAD | SKIP_EMPTY flags of LineReaderTask (no observable effect: SKIP_EMPTY only works with DROP_NEW_LINE, empty lines are emitted, as already tested), and flush calls masked by the repeated flushes (below).

Observed while writing the tests, not changed here (pinned only where it is guaranteed):

  • a FlushableTaskInterface task is flushed several times: once its iterable parent has no more item, then again each time one of its ancestors or itself is resolved (resolve() flushes the subtree of each task); it is also flushed when it has never been executed (empty iterable parent), and after a stop during a flush;
  • the output of a flushable end point is reset by its flush (the process returns null);
  • ProcessState::stop() without error does not flag the process as stopped (ProcessEvent::isProcessStopped()), only the stop task does.

🤖 Generated with Claude Code

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@njoubert-cleverage
njoubert-cleverage requested a review from a team as a code owner November 3, 2026 13:55
@njoubert-cleverage
njoubert-cleverage merged commit f7c33d3 into main Nov 3, 2026
17 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

Tests: behaviors shown untested by the surviving mutants (skip, stop propagation, xpath, base_path, cached ttl, libxml state)

1 participant