summaryrefslogtreecommitdiff
AgeCommit message (Collapse)Author
2023-07-04actually set Build.shaiximeow
2023-07-04set a limit for how far back we'll construct new runs of existing jobs for ↵iximeow
new hosts
2023-07-04parse cpu MHz from /proc/cpuinfo correctlyiximeow
2023-07-04try reporting max cpu freq, but this is going to be messyiximeow
2023-07-04fix host info queryiximeow
2023-07-04adjust metrics to show multi-host summaries?iximeow
2023-07-04cap max wait time for new tasksiximeow
2023-07-04actually run spawned task to get clients jobsiximeow
2023-07-04too noisyiximeow
2023-07-04let clients keep connections alive a little longeriximeow
2023-07-04more sql query typo fixesiximeow
2023-07-04repo row is non-optional, value in `default_run_preference` isiximeow
2023-07-04that query was just wildly wrongiximeow
i dunno how the `and not exists` is going to work out over time, probably need some kinda checkpointing so that i don't have to rule out the entire build history for a host every time it asks for a new task
2023-07-04missing query paramiximeow
2023-07-04fix type'siximeow
2023-07-04pretty sure this gets host-preferential task assignment in place?iximeow
2023-07-03expose function for lua to run commands and collect outputiximeow
onward, `rustc --version`!!
2023-07-02try to support cross-arch runners....????iximeow
2023-07-02fish out host info for ... at least the nanopi m4 on the bookshelfiximeow
2023-07-02use the platform-generic c_char, instead of i8/u8 WHICH I GUESS VARIES ↵iximeow
BETWEEN x86 AND aarch64
2023-07-02actually capture cpu family for ... cpu family ...iximeow
2023-07-02groaniximeow
2023-07-02better sql i thinkiximeow
2023-07-02more cpu_vendor misnamingiximeow
2023-07-02silly deadlock bug, sql typosiximeow
2023-07-02add host identifiers?iximeow
2023-07-02last run for job should be the chronologically most recentiximeow
2023-07-02do not overload the tag "kind"iximeow
task_info kind overwrites the containing clientproto kind
2023-07-02client ack to new task is now kind:started, status:startediximeow
2023-07-02use new protocol bits when ackinga task requestiximeow
2023-07-02add a protocol, host info reportingiximeow
2023-07-02cert path and key path are no longer mandatoryiximeow
2023-07-02make real and debug webserver addresses (and tls-ness) configurableiximeow
2023-07-02make client configurableiximeow
2023-07-02better description for reaped tasksiximeow
2023-07-02cleanup, loggingiximeow
2023-07-02clean up old stale jobs for clients that have gone awayiximeow
2023-07-01produce short sha after looking up the full sha so short urls do not panic ↵iximeow
from invalid bounds
2023-07-01worst testing loliximeow
2023-07-01create a new run to go with a jobiximeow
2023-07-01fix client sending wrong task status nameiximeow
2023-07-01job->run in driveriximeow
2023-07-01again, column skew..iximeow
2023-07-01fix pending jobs query, right number of columnsiximeow
2023-07-01short hashes for og previewiximeow
2023-07-01stray ref to old jobs pathiximeow
2023-07-01add og tags to commit summaries?iximeow
2023-07-01add a notion of runs distinct from jobs, lets see how well this goes overiximeow
2023-06-30specifically report failures to check out build refsiximeow
2023-06-30release db connection lock so as to not deadlock ci_driveriximeow
afaict holding the lock over an await point gives tokio an opportunity to run a different task - perhaps a different API request that would try acquring the db connection lock. then it deadlocks; the second task holds the thread hostage while waiting for the first task to release its lock, which it can't do due to no thread being available for it to run on.