[Pkg-javascript-commits] [sockjs-client] annotated tag upstream/0.3.4 created (now 72e9382)
Tonnerre Lombard
tonnerre-guest at moszumanska.debian.org
Wed Jan 8 00:50:14 UTC 2014
This is an automated email from the git hooks/post-receive script.
tonnerre-guest pushed a change to annotated tag upstream/0.3.4
in repository sockjs-client.
at 72e9382 (tag)
tagging 9b14ec3de19d8b692ed84bf3dee78521efab3e55 (commit)
tagged by Marek Majkowski
on Thu Nov 15 14:29:31 2012 +0000
- Log -----------------------------------------------------------------
Release 0.3.4
Carl Byström (1):
Updated jQuery to v1.7 to prevent event warnings from appearing in WebKit browsers.
Casey Watson (2):
Update README.md
Update README.md
Christian Sturm (1):
There now is a SockJS-perl
Jakub Stastny aka botanicus (1):
Link to the sockjs-ruby repo.
James Halliday (1):
make `new` optional
Marek Majkowski (419):
Initial commit
License boilerplate
Don't hardcode private ip address in tests
Ignore automatically generated files
Test batched messages
Opera and IE jsonp script quirks.
More tests: 404,500 and connection refused tests.
Error on script now works even for konqueror.
Generate prettified javascript, useful for development.
Local references to SockJS renamed to Sock, for better minification - 200 bytes saved.
Sock goes back to SockJS, there is a better way after all.
Use local references to document and window
More powerful generating script.
Simple server shall compress output and support if-modified-since.
Json2 is now bundled with sockjs.js
typo
Allow uglify to rename even top objects.
Support for eventstream transport.
ws transport renamed to websocket.
Htmlfile transport.
Flying cursors example.
Add a landing page.
Dedup iframe code.
Make sockjs_url configurable.
Let chrome/safari not shout for XHR.
Iframe timers fixed.
Update copyright holder.
Heavy refactoring, open is now a protocol, not a transport thing. Close can be initiated only by the client.
Enable passing complex objects.
Send at most 5 events per second in the cursors example.
Smoke test - measure latency in long-lived connections.
Start of the work on xhr polling transport
Refactoring the world continues.
Xhr CORS polling transport.
XHR cross domain - XDR. IE implements it differently.
New firefox has MozWebsocket object
Brief readme
Brief readme
Brief readme
Few smoketests.
Renamed some transports, don't set devel/debug by default
Make jsonp work on ie7
Renamed xhrpolling to xhr-polling, other changes
Htmlfile transport is not supported any more
xhrpolling --> xhr-polling
Print time of generation
Let's use the name XDR for cross domain XHR.
Treat xhr-streaming as a specialized part of xhr-polling.
Opera mini needs to wait a while before we unlink iframe from jsonp sender.
Apparently \x00 doesn't go through xhr in ie. Replaced with \n.
XDR can be cleaned up during the callback, so double check if it isn't already cleaned up.
Improved error reporting in tests.
we do support xhr-streaming now
Link
Allow any protocol for cursors.
Added forgotten whitespace.
Increase the jsonp timeout to 35 seconds.
Don't use setInterval, it's evil.
Play with cacheability, etag instead of last-modified.
There isn't any verbatim socketio code any more.
Tests moved to a separate directory
Examples/tests moved into separated directory, node code included.
Use makefile
main.js --> index.js
Merged two CS files with qunit tests to one.
Recompile on modification (make serve)
Versioning, prepration for the first release
Release 0.0.0
Automate even more
Release 0.0.0-dirty
Few words of comment
There is not much value in releasing --pretty' builds
Few more words
Apparently, the versioning is used in completely different way
Few small improvements
More readme
Javascript does depend on the version file
Few more words in readme
Version encoded in javascript shall be configurable
We don't use package.json for releasing, so the version there doesn't matter, and it's only for npm sake.
typo
Release 0.0.1
Accept urls starting with '//' or '/'
Simplified running qunit tests.
We run a single server now.
Deployed tests
Forgotten file, due to too wide .gitignore.
Set author in the json, even though it's not going to go to npm.
Advertise mailing list
Githubify markup
Do accept cookies with CORS.
Better, more powerful configuration for tests.
Added changelog
Release 0.0.2
Generic htmlfile support
Activex IE version of htmlfile transport
Added second 'transports' table, it's close, but not identical to the first one
Fixed bug in EventEmitter code, added some tests.
New generic infrastructure to handle polling transports. Htmlfile transport refactored.
IE doesn't support [].indexOf.
In order to check if polling transport properly restarts connections, we need to send more data in tests (>128K).
Move EventSource transport to the generic polling infrastructure.
New transport: xhr-streaming, it will duplicate functionality of xhr-polling for a moment.
Don't run a EventEmitter callback if it's set to null
No need for that, polluting is already done in the HtmlfileReceiver constructor.
It should be configurable whether XhrReceiver should use CORS-enabled ajax or not.
IE doesn't like peeking into responseText and status on readyState=3.
Unified all XHR/XDR transports.
Helper to count a size of object.
In some circumstances the xhr during 'cleanup' is already nulled.
Cosmetic.
Messages exchanged between main window and an iframe must be tagged with window_id - in order to identify a proper iframe.
Do pass fourth parameter to the iframe-enabled transports - base_url.
Enable chunking-detection for streaming protocols (with the exception of websockets).
Test a long utf message - to find if server is able to merge multiple packets without breaking utf.
Chunking test shall work on opera.
chunking detection needs to be within the loop. Maybe the whole loop should be made as a recursion some day.
options.cookie may have one of three valid values: true,false or undefined.
We need bool, not a websocket object.
In opera (where xhr doesn't do cors), correctly pass errors back from chunking detection (close events).
Example/smoke tests revamped.
Tests for chunking detection.
This package isn't npm-releasable, so devDependencies make more sense.
Prevent re-running chunking due to timeout.
In tests we use sockjs.chunkingTest with raw url, which may be empty string or slash. Normalize it before using.
In Safari 'EventSource' is an object, not a function. Make es detection more robust.
Wrong url tests take a long time and tend to break. Move them to the back, so they should be run in the end.
EventSource crashes chrome < 15 and safari when window is closed too quickly.
Check if coffee is installed before running command. Prevents producing empty files when node isn't there.
No need to run some tests twice.
Cosmetic, few words about deployment and load balancing.
Check if coffee is available before running it.
Share options with the chunking test - for example 'devel' option affects iframe url (in opera).
cosmetic
Make sure xhr is always aborted, even if user leaves the page.
Better error messages in tests, tweaked large_message test data.
Release 0.0.4
cosmetic
Don't run error in-flow, should be oob.
Renamed sender js.
Delay sending messages by 25 ms, in order to make tests more stable.
cosmetic.
Instead of uploading loads of data, we're in fact insterested mostly about the testing of messages from the server.
Readme tweaks.
Utils.delay
Change directory before playing with npm
addEventListener doesn't check if listener was already added
Merge branch 'master' of github.com:majek/sockjs-client
Default config should be generic, not necessarily simple.
#7 cosmetic, code smell
#7 closed, some tests
Readme header.
Repo migrated
A note on SockJS limitations
Hold on with advertising sockjs-ruby
Add disabled-websocket-test
Merge branch 'master' of github.com:sockjs/sockjs-client
Be less verbose during tests.
Fix #8. Throw exception on missing url.
Fix #13. Automatically convert data to string.
Fix #14. Renamed `status` property to `code` on CloseEvent.
Tests should run with debug=true by default. Why not.
#14 support `wasClean` property on CloseEvent
#14 Forgot to update smoke tests.
Update example to use more idiomatic sockjs-node api.
Idiomatic api now uses `createServer`
sockjs-node API refactoring continues.
Add link to Erlang server
Cosmetic: spotted by @dvv
IE9 fires `onerror` event randomly. Let's delay it for a while to make sure that json really wasn't loaded.
Updated Changelog
Clarify opening multiple SockJS connections using subdomains.
Supply js named like `maj.min`.
Use sockjs from CDN
cloudfront can do ssl
Upload releases to cdn cloudfront
Merge branch 'master' of github.com:sockjs/sockjs-client
fixed s3cmd usage. running cfdelete is a very bad idea.
Use more standard 'text/plain' rather than custom 't' type for xhr
Introduce first go of utf encoding tests.
Merge branch 'master' of github.com:sockjs/sockjs-client
Cosmetic: reformatting
Tell something about utf-8 messup
Release 0.1.0
cosmetic: spurious quotation marks
use 0.1 instead of 0.0
Advertise cross-domain streaming transports
#21 Support the usual jsonp trick on ie9
Fix #18: start documentin browser quirks
Mention single-sockjs-connection limitation in the quirks section
Fix #26 - emit heartbeat event when heartbeat happens.
Fix #27 - include license inline
Changelog
Mention subdomaining possibility.
#15 - mention inability to load SockJS in head as a limitation
#24 for all IE quirks - use feature detection to check if it supports streaming (during tests)
Release 0.1.1
untabify
#29 - Add custom utils.quote and new test infrastructure for that
A bit better implementation of random_number_string
Cosmetic
Deployment: Clean build before tagging
Make sure there are no raw unicode characters in generated minified javascript
Update uglifiy-js
Better tests for utf encoding
Fix #27 - use custom JSON quoting function
Deployment: watch the other directory also
Escaping new lines is not exactly a good idea.
Issue #29 cosmetic: 'Lookup' is a better name
#29 IE doesn't like \ufff0, escape it also.
#29 Explain that we now support all unicode chars.
Removing unused `minify.coffee` script
Run both new unit tests and old tests from the main quinit test landing page
Don't duplicate amendUrl tests.
Preparing to run unit tests on testling.
Remove spurious comment from generated minified js
Cosmetic: indentation
Test all \u20XX chars
Old Firefox has an interesting JSON encoder - encodes tab as \u0009 instead of \t
Move functions requiring dom from utils.js to dom.js
Make the code shorter - use substr instead of array index.
Support new node. (@dvv)
Simplify makefile - coffeescript compilation of tests may be simpler
Framework for dom tests.
Make `make serve` twice faster - compile coffee before forking server process.
Test onunload and onmessage events - using an iframe.
One more test for amendUrl.
Don't use console.log, use log instead. Reload iframe every time (ie: disable caching)
Get document.domain trick/same origin policy work for ie7
Fix for Onunload not being triggered in ie7 and ie8.
Run domtests with the normal quunit tests.
#28 Test onbeforeunload event
Abstract similar onunload and onbeforeunload tests
Actually, timeout of 0 works as well as 10.
Don't start test onload, we want to test onload as well!
iframe can be null in a timer
#15 Some transports require document.body to exist - wait for onload
#15 add tests
Spurious header.
#28 Introduce DocumentGuard, a singleton that knows more than document.readyState.
#28 Don't start a websocket connection if we're after beforeunload
Cosmetic
#28 Emit warning.
Having a global `onmessage` function is a bad idea.
Cosmetic
updated changelog
Release 0.1.2
Dont forget to compile unittests.
Merge branch 'master' into dev
Test server was moved out to sockjs-node
Cosmetic.
Cosmetic.
changelog updates
Fix #34 - Don't reuse `session_id`.
Cosmetic: Add links to `config.js` files.
Don't test Unicode surrogates - it can't work against python sockjs server.
Advertise Lua/Tornado/Vert.x SockJS servers
Merge branch 'master' into dev
Fix #35 - forgot to json stringify/parse
`arrSkip` util function
SockJS-node is not required any more
Cosmetic.
#36 Add info.js, a chunking-test.js replacement.
#36 Remove chunking_test remains
Cosmetic
#36 delay shall be called RTO
#36 It's possible to have empty options.info, during tests.
New infrastructure for AJAX/XHR/XDR plus tests
More stable streaming test (first chunk can be smaller than 2049)
cosmetic.
Small security check for iframe.
Cosmetic - additional null-ness checks
Make info working on Opera (ie: needs an iframe to do cross domain ajax)
Timeout/delay/rto is dependent on number of rtt.
Use new ajax machinery for info, fake info response on ie7
IE doens't have console.log
Fighting with IE
utils2.js --> dom2.js
Abstract creating of cors ajaxe objects
Don't disable websockets during tests (when options.info are empty)
Cosmetic
Move xhr transports to new ajax infrastructure
Drop old ajax creation methods, copy over some comments to new code.
Github changed the markdown format. Is that better?
Markdown: maybe 4 spaces?
Markdown: bold not italic
New transport table.
Markdown: Let's try dagger instead of star. Multiline tables don't work
Markdown: cosmetic, removed old tables
Add a note about disabled websockets to readme.
Cosmetic: _window instead of window
New protocol detection algorithm.
Refactoring of ajax-based transports. All now use a common AjaxBasedTransport class.
Separate xhr-polling from xdr-polling - new name for xdr transports.
Use new detectProtocols method to detect protocols.
Cosmetic: _window instead of window
In sender use passed XHR/XDR object, don't try go guess again.
Drop automatical detection of xdr - createXHR exists no more.
Fix the 2000 error on invalid paths. Now expect 1002.
Move utils object declaration up.
Cosmetic.
IE7 doesn't look at /info, thus returns different error (2000 not 1002)
Cosmetic: Moving tests around
Cosmetic: trans-xhr-streaming.js -> trans-xhr.js
Few more comments about jsonp.
Forgot to mention xdr-* transports in some html examples.
Make it possible to override rtt.
Disabling just websockets is sometimes useful - for example cursors.
Jsonp only shows spinning cursor on send.
Cosmetic: gray not grey (this upsets ie7)
Cosmetic.
cosmetic: mention ie6
For some reason turning off 'disabling' didn't work on ie7.
Drop document_guard. Use onunload instead.
Cosmetic: don't put </script verbatim in the javascript.
API change - handle `protocols_whitelist` option instead of second parameter to the construcotr.
Set an 8 second timeout for xhr info requests.
Cosmetic: there is no fail(true) in qunit - use ok(false) instead
Cosmetic: different message for config disalbed protocols
New test: close on close
Rename smoke-throughput --> smoke-latency
Changelog
Disable onbeforeunload test - it's not used any more.
Actually, xhr transports require preflight - two round trips
Cosmetic
Readme updated to 0.2
Release 0.2.0
I don't get it - makefile
Release 0.2.0
cosmetic
Cosmetic - No need to hit the limit on every send - forgot to remove payload.
Cosmetic: rename isLocalUrl --> isSameOriginUrl
Cosmetic: No need to verify window.ActiveXObject - we're already in try/catch block
add getParentDomain() helper
Bump core dependencies
Apparently \xFFFE and \xFFFF aren't valid unicode characters
Mention problems with file:// urls
Change the semantics of info option a bit - extend not overwrite
Merge pull request #46 from nehresma/master
Update makefile to avoid build problems in the future
Changelog
Merge branch 'master' into dev
Release 0.2.1
Further makefile tweaks
Mention https in readme
Mention other sockjs related projects
Merge pull request #47 from cgbystrom/jquery-patch
Mention SockJS behaviour on browsers blocking third party cookies.
Link to the source about the connections per domain restriction.
Fix #50 - fix bug in the isSameOriginUrl method
Mention raw websocket url in the readme
Mention raw websocket url in the readme
Merge branch 'master' into dev
When clearing form.are in JSONP set it to empty string instead of null - Maple / Samsung doesn't like null value.
Fix #49 - Support for sockjs-client being used from pages served from file:// addresses
#51 - try using XMLHttpRequest instead of XMLHTTP one
#28 - Handle onunload / onbeforeunload more robustly.
Added description to package.json
#51 - IE10 should use xhr before xdr
sockjs/sockjs-protocol#28 - always pack the message in square brackets
Fix #53 - AMD compliance
Try to be more helpful and mention a bug in README.
Refuse SockJS urls with ? or # inside (query strings or hashes)
Verify origin of the agent sending postMessage events
Use different xhr object for local and for CORS ajax.
Fix #55: don't leak iframe object, expose postMessage interface only
Fix #57 - Opera adds iframe navigation to history
Note that non-SSL connection may not work from SSL-served page.
Temporarily disable iframe tests - they are failing too often.
Don't mention heroku, mention https Europe.
Changelog
Release 0.3.0
#57 - another try - don't set the iframe src as it adds an element to history
Let the iframe / htmlfile code use our new unload_add hooks.
#57 - Trigger cleanup also onbeforeunload, don't set iframe.src.
Fix #60 - Opera 12 doesn't do xhr-streaming
Suggest using SockJS 0.3 not 0.2
Merge branch 'master' into dev
Added IE 10, updated RFC6455 implementations
Add license blocks to every file
more ws rfc links
Note on Heroku
Merge branch 'master' into dev
Fix #58 - increase the timeout for iframe
Merge branch 'dev' of github.com:sockjs/sockjs-client into dev
#57 - don't test onunload in opera
Merge branch 'dev' of github.com:sockjs/sockjs-client into dev
#58 - Opening an ajax to a wrong dns results in a very long wait
Fix #63 - delay xhr-streaming until onload
Canceling SockJS before "info" exited caused an exception in IE.
Allow longer connection time for ws (roundtrips = 2)
Cosmetic: don't do alert, do utils.log
Test for #61
Fix #61 - catch all was two lines too wide, whoops
Changelog
Release 0.3.1
advertise sockjs-cyclone
Mention SockJS-twisted
Merge branch 'master' of github.com:sockjs/sockjs-client
Fix #77 - SockJS doesn't work from file:// urls
Release 0.3.2
Release 0.3.2
mentin wai-sockjs
Poke tables
npm --dev doesn't behave
Don't use javascript: links
Merge branch 'master' of github.com:sockjs/sockjs-client
#94 - SockJS is more stable over SSL
#90 - don't even catch onbeforeunload
Work around another IE bug - 204 gets mangled to 1223
it is impossible to cancel JSONP polling request - comprehend that
updates to EventEmitter class (events -> _events, etc)
introducing EventEmitter.on
Mention @njoyce fork of sockjs-gevent
Merge branch 'master' into dev-new
Fix #66: failure to post data to /xhr_send should kill the session
Cosmetic - Use local copy of window object
Cosmetic - sockjs works better over ssl
Release 0.3.4
Michael Bridgen (1):
Use the npm-installed coffee, rather than relying on a global coffee
Nathan Ehresman (1):
Typo fix
Vladimir Dronnikov (1):
strip trailing slashes in URL, to avoid possible double slashes after URL concatenation
majek (3):
Edited README.md via GitHub
Use new google groups interface
Edited README.md via GitHub
-----------------------------------------------------------------------
No new revisions were added by this update.
--
Alioth's /usr/local/bin/git-commit-notice on /srv/git.debian.org/git/pkg-javascript/sockjs-client.git
More information about the Pkg-javascript-commits
mailing list