[qt/qt-creator/qt-creator]: Summary of bulk changes made
KDE Git Services - Bulk Change <[email protected]>
| Newsgroups | gmane.comp.kde.cvs |
|---|---|
| Message-ID | <[email protected]> |
Git repository change summary for qt/qt-creator/qt-creator Pushed by mirror-service into branch 'master'. Changed from cb786b2fab5302bb1a7d646a752583c6116c6af9 to 785c71422877d93e1adb70f0a5d198d505ec394b Acknowledgement was received that this change introduces only existing code that has been pushed to another public open source repository. This change contains the following new commits: Git commit b5f04d71bbf83b7c845ab9888831728cf1648949 by Alessandro Portale on 13/08/2026 at 12:16.. Tracing: Avoid loop-detach Change-Id: Ic15cad095316c76a30c496c5a2be18884b7b2109 Reviewed-by: Marcus Tillmanns <[email protected]> https://invent.kde.org/qt/qt-creator/qt-creator/-/commit/b5f04d71bbf83b7c845ab9888831728cf1648949 Git commit c505f607ea75ee4cf9444cbd230662d932f73e9a by hjk on 13/08/2026 at 13:59.. McpServer: Let a key query target a window press_keys and type_text rejected a window-level query - e.g. window_title alone, which matches every widget in the window - as ambiguous. For a key the intended target is that window's focused widget ("press Enter in the About dialog"), so resolve such a query to the window's focus widget via a new resolveKeyTarget; it still errors when the matches span several windows. Assisted-by: Claude Code Change-Id: I64d24bedaf31c6ffc3f685b55ce7d3a6827cdaf3 Reviewed-by: Alessandro Portale <[email protected]> https://invent.kde.org/qt/qt-creator/qt-creator/-/commit/c505f607ea75ee4cf9444cbd230662d932f73e9a Git commit d446dd72e6e55b539b796e837692eba793f2218b by hjk on 13/08/2026 at 13:59.. McpServer: Name scenario keys after the MCP tools The scenario action keys diverged from the MCP tool names (click vs click_widget, capture vs screenshot, invoke_action vs call_action). Rename the keys to match: call_action, click_widget, type_text, press_keys, select_combo_item, screenshot. The composite/assertion keys stay - menu drives find_menu_item + activate_menu_item, and expect/expect_gone/wait_for are widget_exists. Tool names are left as-is: descriptive names help an MCP client pick the right tool among many. Assisted-by: Claude Code Change-Id: I17dd0cbc581c461121efe82a9f018b4422e32cc1 Reviewed-by: Alessandro Portale <[email protected]> https://invent.kde.org/qt/qt-creator/qt-creator/-/commit/d446dd72e6e55b539b796e837692eba793f2218b Git commit e64128ab71981bda6d920f719554c615761486f0 by hjk on 13/08/2026 at 13:59.. McpServer: Match a widget by its buddy label's text The widget tools already match on a widget's own visible text (button, label, group box, combo box). An input like a line edit has no such text, so also match the text of the QLabel whose buddy() points at the widget - addressing "the field labelled Search for:". find_widgets reports that buddy text too. A label that captions another widget (has a buddy) is not matched by its own text, so such a query resolves to the field rather than its caption. Own-text matching is unchanged for plain labels. Assisted-by: Claude Code Change-Id: I67c35929ad81f8ac9e1a0f624658b31bb7f01b1f Reviewed-by: Alessandro Portale <[email protected]> https://invent.kde.org/qt/qt-creator/qt-creator/-/commit/e64128ab71981bda6d920f719554c615761486f0 Git commit 785c71422877d93e1adb70f0a5d198d505ec394b by Jeff Heller on 13/08/2026 at 14:28.. ProjectExplorer: Do not report outdated issues over MCP The issue tools reported whatever was in the Issues panel, including code model diagnostics describing a file as it looked at the last parse. The QML code model only re-parses on project events, never on a build or on a file changed outside Qt Creator, so an error could outlive the text that caused it: the build passes, qmllint is clean, and the tool still reports a syntax error on a line that is fine. Acting on that means editing correct code. Remember the modification time of a task's file when the task arrives and leave the task out once the file differs. Build tasks are exempt: BuildManager clears them at the start of every build, so they always describe the latest one, and dropping one would turn a failed build into a failure with no issues to explain it. Also report the task category, so a caller can tell a compile error from a code model diagnostic that no build will ever clear. The same reasoning applies to build()'s verdict, which drew its error and warning counts from every tracked task: application output or a code model diagnostic could report an error against a build that produced none. Count and list only build-category tasks there. run_project and debug_project keep the unfiltered view, because they deploy before running and a deploy failure must stay visible. Tracking a QDateTime per task made a pre-existing hazard fire: the IssuesManager statics were destroyed at exit(), after the owning library had released its memory, and the extra shared-data deref turned that into a crash on quit. Leak all three, as anything living for the whole process should be - there is nothing to reclaim, and running the destructor is what breaks. Co-Authored-By: Claude Opus 4.8 <[email protected]> Change-Id: If9c14deb820d80daf8d56b6be884289a0ad92d5c Reviewed-by: Jeff Heller <[email protected]> Reviewed-by: hjk <[email protected]> https://invent.kde.org/qt/qt-creator/qt-creator/-/commit/785c71422877d93e1adb70f0a5d198d505ec394b