* Add namespace labels to test framework diff snapshots
* fix lint
* failing tests
* more failed tests
* more failures
* more fixes
* one more
* fix fault injection
* Master has a non release gateway API. Changes to allow that.
* Grab a istio test ref from main branch
* <ove from istio.io/pkg to istio/istio/pkg
* revert one change
* update istio/api
* Run make gen
* disable failing test temporarily
* Fix lint errors with new build tools
* Move copyright so it doesn't get marked as a heading
I beleive Istio removed copyrights so we could do a bigger clense
later.
* Update to use the master branch of istio.io/istio for test refs
* go.* changes
* Update test and go.*
* Update to use `master` branch for make targets
* One final go mod tidy
* REmove vm test for now.
* Remove istioctl-analyze test
* Also remove using-istio-dashboard
* Remove file mount egress documentation
This is actively leading users down a bad practice. We previously did
the same for Ingress - the results were we got a lot less bugs about
file mount being very hard to use.
As is, users are directed here as the default - only if they happen to
know what "SDS" is (an implementation detail) will they realize the
other doc is better.
* gen snips
* fix test
* Fix inject
Overview of the changes:
- Adding ability to verify that expected output occurs a number of times consecutively. This is needed for https://github.com/istio/istio.io/pull/8402.
- Moving snapshot checking logic to Go code so that it can be separated out into separate test steps, which are timed and contain their own output directories/files. This makes the code cleaner and also makes the snapshot logic more transparent.
- Updating debug.sh to use newer bash syntax that allows it to dynamically select a free file descriptor. Without this, I was seeing all commands echoed to my console in goland.
This required some other changes WRT verification:
- Change __cmp_like to allow for not accepting <pending> for an IP address.
- Change __verify_with_retry to use a timeout rathan than number of retries. This is a more intuitive interface and aligns with the way we do retries in istio/istio. I also got rid of exponential backoff and allow both the timeout and delay between retries to be configured.
This was noticed when investigating why errors are not causing the test to fail. It's not clear if this is related to the problem, but the existing logic was definitely atypical.
* test framework enhancement for multicluster doc test
* simplify multicluster test framework with setup=multicluster directive
* restore documentation comment
* Update istio version for testing
* Framework changes in base istio required changes
* Work on whitelist
* Remove whitelist and move to a new feature
* Another istio.io/istio update
* Another framework change - remove test suite name
When tests are run, shell script debugging is enabled and written to a
file called "<script_name>_debug.txt" in the same directory as the
"<script_name>_output.txt" file.