Tags: feldera/feldera
Tags
[web-console] Brighten source-range highlight in profiler SQL view Closes #6404. Clicking a node with a source position in the profiler diagram calls Monaco's `setSelections` to highlight the matching SQL range, but the editor is read-only and never focused, so Monaco rendered the selection with its faint "inactive" colour (vs: #E5EBF1, vs-dark: #3A3D41) — the user could barely tell anything had been selected. Register the same `feldera-light` / `feldera-dark` Monaco themes the pipeline editor (`CodeEditor.svelte`) already uses: they override `editor.inactiveSelectionBackground` to a saturated blue with substantial alpha. Defining the themes again here is safe — `monaco.editor.defineTheme` is idempotent and the values match — and unit tests pin both halves of the contract. Co-Authored-By: Claude Opus 4.7 <[email protected]>
pipeline-manager: add `use_platform_compiler` program config option When `runtime_version` is overridden, the pipeline manager normally downloads the matching SQL compiler JAR for that version. The new `use_platform_compiler: bool` field in `ProgramConfig` (default: false) skips this download and uses the platform's SQL compiler JAR instead. This is useful when network access is unavailable or slow, or when testing a runtime version that is compatible with the platform compiler. The Rust runtime sources are still checked out and compiled from the requested `runtime_version`; only the SQL-to-DBSP compiler JAR selection is affected. Written by claude. Tested by me with a local cloud deployment. Signed-off-by: Ben Pfaff <[email protected]>
[adapters] Ignore clock for pipeline completion when now_http_driven. When a pipeline uses NOW, then normally it cannot ever signal that the pipeline is completed, because the results of the pipeline can always change as time goes on. However, when the user has configured the clock to change only as driven by explicit requests, this is no longer the case, so we then can ignore the clock for the purpose of pipeline_complete. Related: #5076 Related: #4340 Signed-off-by: Ben Pfaff <[email protected]>
pipeline-manager: rust compilation cleans up `target` The Rust compilation cleanup now checks whether the disk on which it takes place it about to reach a usage threshold: - If it exceeds 70%, a warning is printed. - If it exceeds 90%, the `target` directory is removed if it is 5% or more of the total disk space. The compiler server is restarted if the form factor supports it. The OSS edition form factor does not support this restart behavior, and as such it will mean that it will need to recompile from scratch. A warning is printed to inform of this, as an indication to increase the disk size. This behavior is currently gated behind the unstable feature `rust_compiler_full_cleanup`, and thus needs to be explicitly activated. Signed-off-by: Simon Kassing <[email protected]>
docker: update go to 1.26.4 we are updating to fix CVE-2026-42504 Signed-off-by: Swanand Mulay <[email protected]>
python: fix long test names This seems to be a problem for k8s which fails with metadata.labels: Invalid value: "bf920_test_cannot_edit_runtime_config_resources_namespace_if_not_cleared": must be no more than 63 characters In the future we may need to restrict pipeline names directly rather than just making sure tests don't exceed it. Signed-off-by: Gerd Zellweger <[email protected]>
Fix cdc_order_by with multiple fields. PR #6245 caused a regression for `cdc_order_by` which prevented parsing of order by clauses with multiple fields. That was because we didn't use the right DF function to parse it. This corrects the mistakes and adds a series of tests. Signed-off-by: Gerd Zellweger <[email protected]>
PreviousNext