Fixed

• Reports for Submit no longer escape HTML entities (!180)

Added

• Submit script API integration (!177)

Changed

• PyLint is now an external dependency (!177)

Fixed

• Some analyzers were not loaded (!176)

Added

• Improved worker management: viewing logs is a single click, just like canceling or outright purging the worker. (!168)
• An in-worker watchdog that terminates the worker if no ping response arrives within a set interval. (!168)
• Submissions indicate whether their job was canceled. (!168)
• Users can request the datetime their current token will expire. (!168)
• An API endpoint, /api/v1/worker/events, that sends server-sent events with interesting happenings around workers. The UI uses these to update the Workers page, instead of hammering the getWorkers endpoint slightly faster than twice per second. (!168)

Changed

• Exceptions raised within jobs now include the full stack trace and not just the message. (!168)
• The checker routine no longer swallows analyzer exceptions. (!168)
• The worker Python detector now also indicates the path to the found Python executables. (!168)
• Worker shutdowns through messages now happen within three seconds (down from two minutes). (!168)
• The CORS patch filter now also fires if Access-Control-Request-Headers is not present. (!168)
• Killing all workers now happens in parallel to speed up shutdown. (!168)
• Worker tokens expire only after a full century (!173)

Fixed

• Concurrent logging on a worker that was causing crashes extremely rarely. (!168)

Description

Adds a memory limit 4 GiB to Docker containers.

Changes

N/A

Additions

• Allocate at most 4 GiB of RAM (including swap) to a Docker container
• Set the name of the docker image to the submission ID for easier identification

Changes

• Fix the REST exception handler sending content types including stars (*)
• Fix the assignment controller erasing existing timeout values if the PUT request did not contain a timeout
  • The old behavior is arguably what the HTTP spec says should happen, but aside from hand-crafting requests to set the timeout correctly there is no good way to preserve timeouts with the current UI

Additions

• Send a kill command if the worker has had the same job for longer than the timeout allowed

Major changes:

• Benchmarking and histogram views
• Workers generate more fine-grained results using multi-level entities
• Test file detection
• New UI
• Report generation overhaul
• Removed template support from the API
• Periodic system security checks
• Encrypted worker communication
• Additional metrics and tool integrations
• Prometheus support

Resolves #137 #134 #128 #133 #66 #115 #153 #151 #65 #122 #102 #120 #127 #109 #147 #161 #143 #140 #166 #146 #167 #159 #163 #164 #141 #148 #97 #82 #139 #160 #175 #94 #176 #162 #179 #104 #106 #168 #178 #142 #180 #181 #154 #182