* update Test Runner to handle API errs for non-private test limits
* update record errors to factor in private vs regular tests
- manually reconcile and update all snapshots —______________—
Co-authored-by: Brian Mann <brian.mann86@gmail.com>
* tighten up potential xvfb display problems
- improve the warning message on possible display problems
- reuse existing helper functions for platform detection
- instead of relying on timing mechanisms, if a possible display
problem is detected on linux - capture all of the stderr silently and
if the broken gtk message is detected, retry running
* consolidated and simplified smoke testing, xvfb error logging and messaging
- added —dev support to cypress verify cli command
* fix lots of tests, restore one error message
* fixed remaining CLI tests
* console.log pong only in electron mode
* remove dead code
* retry requests, basic e2e test
* basic e2e test for chrome behavior
* don't use timeouts to test
* some minor cleanup
* validate google chrome's retry behavior w/ a proxy
* get retries on network errors workin
* cleanup
* final changes before switching approach
* Reverting previous approach in request.coffee, server.coffee
* add retryOnNetworkFailure
* now works with retryOnStatusCodeFailure
* retry 4 times in addition to the 1 initial attempt
* add tests for subresources
* much improved error handling
* have the e2e test really work
* e2e baseurl check
* retry baseurl check
* remove special handling for node 8.2.1 bug
* WIP: continue making progress building out request retry
- swap out passthrough’s for pumpify / duplexify / pump
- clean up error handling / messaging / retry logic
* pipe clientrequest events
* buffer req body to disk, restore error handling/retrying
* don't bubble up errors from failed attempts
* actually pipe reqstream, oops
* add some e2e tests for request body
* revert lib/request.coffee to 7be1051
* add almost-working lazy-stream
* manually fire the 'pipe' event on the reqStream to copy headers to the outgoing message
- restore the ‘error’ propagation so that all tests pass for now
* cleanup leaking 'undefined' into stdout, causing failing e2e tests
- do not set onWarning as a property of project, just pass as an
argument
* add new options to request_spec, deduplicate default opts
* use stream buffer in request.coffee
* revert request.coffee
* last stream_buffer changes before removing fs stuff
* remove fs stuff from stream_buffer, add stream piping tests
* it works! :tada::tada:🎉 using duplexify to separate delayStream and reqBodyBuffer
* retry for .5s max on ECONNREFUSED
* add error codes
* don't timeout proxied requests
* restore baseurl check
* update new e2e tests
* make delay work with rp
* propagate clientresponse events
* removing tests that don't do anything now that we don't ESOCKETTIMEOUT on proxied requests
* add new visit, req options to index.d.ts
* don't fail on server-performance-test
* make retries with status code work again
* account for different stack trace in ci
* fix test
* retry https requests
* add tests for https passthru retries working
* clean up error handling for https-proxy
* fix failing https-proxy tests, tweak agent error handling to prevent multiple callbacks
* make expectedExitCode actual vs. expected in the correct order
* bump up e2e test timeout so it can retry and still work
* don't show cypress error page on network errors, gzip errors
* update tests
* clean up onWarning code
* add tests for incompatible options errors
* retry up to 500ms on proxied requests
* remove .only
* add tests for incompatible options errors
* remove .only
* maybe this will help it act more consistently
* help e2e test work in ci
* don't re-pipe the gzip response to client, the response is already gone
In the event that a gzip response has an error and we're not doing full injecting (ie, not buffering the response in memory), we can't just re-stream it.
By the time we do the second .pipe, the stream is already empty.
So, let's just error out anyways, which is what the browser would do (error).
* remove test for injecting into net error 500 pages in iframes
* update visit spec to expect an actual error, not Cypress HTML on visit timeouts
* update el snapshot
* don't reconnect on already made connections
* clarify naming
* wip: testing https proxy
* better debug calls
* WIP: getting proxy passthrough retry tests going
* handle retrying upstream proxy errors
- add tests for successfully retrying proxy errors and for unsuccessful
retries
- fix onClose errors when proxy connection is closed before enough data
is received
- fix not returning setTimeout correctly
* group related code accordingly
* do not build typescript by default, drop extension from main
* more TODO notes
* don't set a default NO_PROXY if NO_PROXY = ''
* debugging-proxy@2.0.0
* null out reqBodyBuffer when req finishes
* wip: retry in agent, not https-proxy [skip-ci]
* update https-proxy to use @packages/network retries
* retry after connection but before proxy tunnel established
* appease my linty overlords
* update https-proxy tests
* update agent specs, decided to still use tls.connect
it's easier to test and has less complexity
* test retrying HTTPS passthru
* debugging-proxy@2.0.1
* increase defaultCommandTimeout 100 -> 200 to prevent flake in CI
* auto formatting
* fix test to be dynamic and not rely on magic constants
* copy types field when linking proxy images, update packages/network types field
* linting
* add network index.js file
* linting
* improve error messaging experience when verifying base url
* only insert 1 new line
* fix failing test not binding to localhost
* removed test that's covered by e2e specs
* remove dash in 're-try'
* some cleanup for readability
* use allocUnsafe per perf
* unset NO_PROXY with an empty string
* move retry ensuring the baseUrl into url, cleanup some imperative code
* if the head is already present during connect, make connection, else wait for first head bytes
* minor formatting, clarity around conditions, naming
* rename retryInterval -> retryIntervals
* set defaults for requests during creation for clarity
* rename send -> sendPromise to pair better with sendStream
* use retryIntervals instead of juggling MAX_REQUEST_RETRIES
- ensure debug messages are consistent between request streams +
promises
- set static constants
* DRY up status check + network failure + retry logic
- keeps the debug logic identical between promises + streams
- ensures all logic paths are also consistent
- consolidates the pop’ing of intervals in a single place
* find / replace fail
* derp
* make the logic actually correct, set intervals as cloned defaults for recursive lookup
* pass arg correctly
* reduce debugging noise, formatting
* rename intervals -> delaysRemaining for clarity
* added unit tests around getDelayForRetry
* set retryIntervals as default options correctly, add unit tests
* using single line when debugging objects
* restore snapshot to match develop
* restore snapshot and move failing tests in e2e visit around momentarily...
* WIP: add the failing tests back in
* add tests for promise/stream that we don't retry on timeouts but do retry on resets
* use create entrypoint for tests
* pass requestTimeout from cy.visit(), remove hard coded request timeout
* s/requestTimeout/responseTimeout
* clean up how we initiate the request stream
* buffer resolve:url response before resolving promise
* only persist the buffer to memory once response finishes
* update visit spec to fail with ESOCKETTIMEDOUT
* use stream.PassThrough to buffer the response stream
* always buffer response stream
* de-commit those screenshots
* buffer forever - node's default is only to buffer 16kB
* some cleanup and renaming vars
* re-remove screenshots
* do all resolution login in on(end)
* update test to call the sendRequest callback
* update comment
* enable resolving:url to cancel currently in flight promises and abort pending requests
* add test for only having one request in flight at a time
* test currentPromisePhase too
* make tests fast again™️
* cleanup the coffeescript conversion, use common-tags, and use res.end to write final chunk
* add test ensuring that responseTimeout is passed correctly when resolving the url via cy.visit()
* micro-opt to check if err.code is property of zlib.constants
* always destroy the request socket when gzip fails as opposed to forwarding the broken bytes
* cleanup, formatting
Co-authored-by: Brian Mann <brian.mann86@gmail.com>
Co-authored-by: Jennifer Shehane <jennifer@cypress.io>
* fix(deps): update method-override to 3.0.0 🌟
* remove unused method-override devDep from driver
* oh, where is my mind, I got these backwards. Remove from server - leave in driver 😑
Co-authored-by: Renovate Bot <bot@renovateapp.com>
Co-authored-by: Jennifer Shehane <jennifer@cypress.io>
* Update Cypress.log to throw when not provided an object as its argument
* Add failing e2e test for Cypress.log being used without a cy command
* Write code to handle situation when log is run outside of a command
- add snapshot for passing situation
* have options be obj by default
* Replace end to end test with unit integration test
* Fix issue where if Cypress.Log is not passed a name, then the Command component wouldn't render
* remove excess whitespace
* Clarifies error when focused element not found
Fixes#851 -> Updates Error Message for Clarity when unable to find a focused element
* Revert change to error message
* Fix typos ("Oop" > "Oops")
* add failing test
* launcher: escape path to browser in getVersionString
* fix test
* Use execa.stdout so we can just pass the path as-is, no escaping
* retry requests, basic e2e test
* basic e2e test for chrome behavior
* don't use timeouts to test
* some minor cleanup
* validate google chrome's retry behavior w/ a proxy
* get retries on network errors workin
* cleanup
* final changes before switching approach
* Reverting previous approach in request.coffee, server.coffee
* add retryOnNetworkFailure
* now works with retryOnStatusCodeFailure
* retry 4 times in addition to the 1 initial attempt
* add tests for subresources
* much improved error handling
* have the e2e test really work
* e2e baseurl check
* retry baseurl check
* remove special handling for node 8.2.1 bug
* WIP: continue making progress building out request retry
- swap out passthrough’s for pumpify / duplexify / pump
- clean up error handling / messaging / retry logic
* pipe clientrequest events
* buffer req body to disk, restore error handling/retrying
* don't bubble up errors from failed attempts
* actually pipe reqstream, oops
* add some e2e tests for request body
* revert lib/request.coffee to 7be1051
* add almost-working lazy-stream
* manually fire the 'pipe' event on the reqStream to copy headers to the outgoing message
- restore the ‘error’ propagation so that all tests pass for now
* cleanup leaking 'undefined' into stdout, causing failing e2e tests
- do not set onWarning as a property of project, just pass as an
argument
* add new options to request_spec, deduplicate default opts
* use stream buffer in request.coffee
* revert request.coffee
* last stream_buffer changes before removing fs stuff
* remove fs stuff from stream_buffer, add stream piping tests
* it works! :tada::tada:🎉 using duplexify to separate delayStream and reqBodyBuffer
* retry for .5s max on ECONNREFUSED
* add error codes
* don't timeout proxied requests
* restore baseurl check
* update new e2e tests
* make delay work with rp
* propagate clientresponse events
* removing tests that don't do anything now that we don't ESOCKETTIMEOUT on proxied requests
* add new visit, req options to index.d.ts
* don't fail on server-performance-test
* make retries with status code work again
* account for different stack trace in ci
* fix test
* retry https requests
* add tests for https passthru retries working
* clean up error handling for https-proxy
* fix failing https-proxy tests, tweak agent error handling to prevent multiple callbacks
* make expectedExitCode actual vs. expected in the correct order
* bump up e2e test timeout so it can retry and still work
* update tests
* retry up to 500ms on proxied requests
* add tests for incompatible options errors
* remove .only
* maybe this will help it act more consistently
* help e2e test work in ci
* don't reconnect on already made connections
* clarify naming
* wip: testing https proxy
* better debug calls
* WIP: getting proxy passthrough retry tests going
* handle retrying upstream proxy errors
- add tests for successfully retrying proxy errors and for unsuccessful
retries
- fix onClose errors when proxy connection is closed before enough data
is received
- fix not returning setTimeout correctly
* group related code accordingly
* do not build typescript by default, drop extension from main
* more TODO notes
* don't set a default NO_PROXY if NO_PROXY = ''
* debugging-proxy@2.0.0
* null out reqBodyBuffer when req finishes
* wip: retry in agent, not https-proxy [skip-ci]
* update https-proxy to use @packages/network retries
* retry after connection but before proxy tunnel established
* appease my linty overlords
* update https-proxy tests
* update agent specs, decided to still use tls.connect
it's easier to test and has less complexity
* test retrying HTTPS passthru
* debugging-proxy@2.0.1
* increase defaultCommandTimeout 100 -> 200 to prevent flake in CI
* auto formatting
* fix test to be dynamic and not rely on magic constants
* copy types field when linking proxy images, update packages/network types field
* linting
* add network index.js file
* linting
* improve error messaging experience when verifying base url
* only insert 1 new line
* fix failing test not binding to localhost
* removed test that's covered by e2e specs
* remove dash in 're-try'
* some cleanup for readability
* use allocUnsafe per perf
* unset NO_PROXY with an empty string
* move retry ensuring the baseUrl into url, cleanup some imperative code
* if the head is already present during connect, make connection, else wait for first head bytes
* minor formatting, clarity around conditions, naming
* rename retryInterval -> retryIntervals
* set defaults for requests during creation for clarity
* rename send -> sendPromise to pair better with sendStream
* use retryIntervals instead of juggling MAX_REQUEST_RETRIES
- ensure debug messages are consistent between request streams +
promises
- set static constants
* DRY up status check + network failure + retry logic
- keeps the debug logic identical between promises + streams
- ensures all logic paths are also consistent
- consolidates the pop’ing of intervals in a single place
* find / replace fail
* derp
* make the logic actually correct, set intervals as cloned defaults for recursive lookup
* pass arg correctly
* reduce debugging noise, formatting
* rename intervals -> delaysRemaining for clarity
* added unit tests around getDelayForRetry
* set retryIntervals as default options correctly, add unit tests
Co-authored-by: Brian Mann <brian.mann86@gmail.com>
Co-authored-by: Jennifer Shehane <jennifer@cypress.io>
Co-authored-by: Gleb Bahmutov <gleb.bahmutov@gmail.com>
* cli: debug explanation for XVFB
* linting
* add chai-as-promised to CLI dev
* show Linux specific error solution if cannot verify
* add todo
* chore: consolidate github issue url logic
* linting
* add npm script lint-changed to quickly eslint fix changes JS files
* retry verify with our XVFB
* update errors and tests
* update CLI tests
* add test for display error message
* fix unit test
* add successful test with retry
* finish verify retry test
* warn users if hit display problem on first verify
* try to detect display problem when running electron and retry with our xvfb
* add warning message to spawn when attempting xvfb re-run
* add test for display retry behavior on spawn
* more comments for clarity
* fix typo
* address #3069 by using chokidar fs events
* use chokidar directly for watching specs and add tests
the watchers lib watches for file contents changing and doesn’t watch for files being added or removed
* fix(deps): update supports-color to 6.1.0 🌟
* bump cli down to 5.5.0 to continue support for Node 4
Co-authored-by: Renovate Bot <bot@renovateapp.com>
Co-authored-by: Jennifer Shehane <jennifer@cypress.io>
