- Aug 17, 2016
-
-
Markus Thoemmes authored
Also introduces warnings on unused imports.
-
Christian Bickel authored
Remove unused variables from WhiskConfig and add the ability to Config to ask for environment-variables.
-
Christian Bickel authored
Fixes #1061.
-
- Aug 16, 2016
-
-
Markus Thoemmes authored
-
Markus Thoemmes authored
-
Mark Deuser authored
Remove `--full` argument from `wsk action list` and `wsk package list` Remove `--service_guid` argument from `wsk package create|update` Add `--summary` argument to `wsk rule get` and `wsk trigger get` commands Add `--summary` test cases
-
dubeejw authored
- Use the Go CLI to install the default catalog instead of the Python CLI
-
- Aug 15, 2016
-
-
NICHOLAS W. SPEETER authored
-
Dominik Jall authored
add null check for logger.
-
- Aug 13, 2016
-
-
Nick Mitchell authored
Update tests in response to critique: remove "enable" from WskRule, pull the rule validation tests out of the withCleaner scope.
-
Justin Berstler authored
-
- Aug 12, 2016
-
-
James Thomas authored
Weather service used by the forecast action has recently released a new API version. This has broken the current Action source, which used the old URL structure. Updating the code to use the new URL paths and parameters.
-
Christian Bickel authored
Make emit implicit instead of passing it around. Use akka Loglevel instead of own Loglevels. Move the logmarker to the end of the message fixes #487. Signed-off-by:
Markus Thoemmes <markus.thoemmes@de.ibm.com>
-
- Aug 10, 2016
-
-
Justin Berstler authored
If a callback is supplied as the "next" parameter, these functions return nothing and uses the callback as before. If no "next" parameter is supplied, these functions return a promise. Added unit tests to ensure that the returned promise behaves as expected, as well as the existing callback functionality still works. Refactor post() to be totally common for both trigger() and invoke() post() now itself returns a Promise
-
Perry authored
Change KVReporter's input thunk so that it accepts a count. Use count to reduce the rate at which user activation counts are written out
-
- Aug 09, 2016
-
-
dubeejw authored
-
- Aug 05, 2016
-
-
Vincent Hou authored
Since services:registrator is removed, we need to remove from the gradle setting. If it is still there, import openwhisk project as a gradle project will raise an error in eclipse.
-
- Aug 04, 2016
-
-
Rodric Rabbah authored
Change units from Int to ByteSize for byte-sized-limited, update comments and json serdes, add byte size subtraction, update tests. Make some const values core package visible. Make memory option byte-sizes. Add toString on byte-size and update error messages. Remove printlns. Update and move tests around between packages. Make TimeLimit constructor accept only a duration. Revert change to log/memory limit constructor so that the type is int not bytesize (really would want to constrain this to be MB but the type conversions aren't that rich and arguably not worth it as this can be hidden behind the private constructor). Change Bytes to B in toString to avoid 1 bytes vs 1 byte. Should it be byte[s] and append the s only when necessary? Emit parameters should test fail. Use python CLI for http test as go cli assumes https. Add some more tests for limits. Use Python CLI for two tests that do not work with Go CLI. Add retry on list operations which might need some time to be consistent.
-
- Aug 03, 2016
-
-
dubeejw authored
- Display more information about invalid arguments
-
Rodric Rabbah authored
Fix link to db README. Remove deadcode. Move gradle helper to gradle directory and update all path.
-
dubeejw authored
- Improve error handling when invalid arguments are passed to action commands - Display an error when an invalid kind is passed when creating an action - Add --kind to action update
-
Rodric Rabbah authored
Modify activation feed so that we grab next batch of messages and commit offsets immediately, essentially marking the activation as having satisfied "at most once" semantics (this is the point at which the activation is considered started); if the commit fails, then messages peeked are peeked again on the next poll. While the commit is synchronous and will block until it completes, at steady state with enough buffering (i.e., maxPipelineDepth > maxPeek), the latency of the commit should be masked. Add kafka test where we intentionally fail the commit and then confirm recovery. (Unrelated by stashed in this commit: Ignore any db local ini in ansible directory in case there are more than one.)
-
- Aug 02, 2016
-
-
Carlos Santana authored
-
Philippe Suter authored
Those have become disruptive on Travis CI. Disabling until we can at least reproduce locally.
-
Philippe Suter authored
Eliminates code duplication, and ensures all systems are properly shutdown.
-
- Aug 01, 2016
-
-
Markus Thoemmes authored
- Making gradle files even more standardized - Moving controller/invoker command to Dockerfile/ansible playbook
-
NICHOLAS W. SPEETER authored
Updated docs for weather name change Updated weather service name
-
Ioana Baldini authored
Add compilation for init with check for compilation errors and modify test accordingly (fixes #493). Add test for bug fix related to unsafe parameter passing ({x: null} for example) Limit stack trace to exclude proxy. Add CLI test for #493. Remove println in test.
-
Evan Patterson authored
The line is incorrect because JSON is not valid Python, e.g., `null` is not defined in Python.
-
Rodric Rabbah authored
Change image for pre-warmed containers from NodeJs to NodeJs6 since the latter is the default for Node actions now. Fixes #976.
-
- Jul 30, 2016
-
-
Mark Deuser authored
- --apihost argument still overrides configured apihost value (including the default value) - Add test case to check for default property values - Fix bug: `wsk property get --apihost` would return all property values Add warning comment to CLI property configuration file backup/restore methods.
-
Perry authored
Send docker pull operation to a different guard so that pull operations are serialized among themselves but can run concurrently with non-pull operations.
-
Philippe Suter authored
The test introduces an actor that acts as a "configurable" HTTP proxy (i.e. can simulate failures). Removing DbPing.scala as the new test subsumes it.
-
- Jul 28, 2016
-
-
Markus Thoemmes authored
Adds the logsize as a limit that is settable per action and thus included in the quota for each action. Document logsize. Handle limit flags as pointers to get proper null values. Adding ActionLimit permutation tests. Signed-off-by:
Christian Bickel <cbickel@de.ibm.com>
-
- Jul 27, 2016
-
-
Fatih Ulusoy authored
Ignoring all swift3 tests because Swift3 is experimental. Some of tests are failed sometimes and break the CI pipeline.
-
- Jul 26, 2016
-
-
Jason Peterson authored
-
- Jul 25, 2016
-
-
Jason Peterson authored
-
Rodric Rabbah authored
Fixes issue #917: do not accept activation id longer than 32 chars if '-' is missing. Add tests for too short, too long and general malformed. Also in this commit: * Tolerate a JsNumber when deserializing an activation id although should never happen. * Remove unused property 'usePythonCLI' since all specialization is done in the constructor. * Relay deserialization error to the client for bad activation ids.
-
Fatih Ulusoy authored
-
- Jul 24, 2016
-
-
Markus Thoemmes authored
This limits the size of logs read from an action container/written to the database by truncating the logs read from the container to the allowed size. Standard (and only allowed value currently) is 10 megabytes. Signed-off-by:
Christian Bickel <cbickel@de.ibm.com> Better documentation Updating documentation bits Moving formatLog method, some refactoring Passing LogLimit instead of ObjectSize Rewrite asserts to withClues Fixed testcases, proper serdes for ActionLimits Adding clause to exit log preparation early Refactor parsing logic, bury size check as deep and early as possible Refactored log parsing, more readability
-