- Work on a pages that modify
Array.prototype(@SevInf).
- Works on Windows again (@SevInf).
- Add
sendFileaction which now should be used instead ofsendKeysto set a file toinput[type=file]elements (@SevInf). - Correctly parse box-shadow in IE9 (@scf2k).
- Fail only single test if reference image is not found (@SevInf).
- Coverage generator now follows symlinks while resolving the paths to CSS files. So the final report will have real paths but not pointing to symlinks.
- Move the coverage HTML templating code to separate module (gemini-coverage).
- Detailed error message when capture region exceeds screenshot area. Such error also will now fail only single state instead of a whole testing process (@SevInf).
- Ignore ``@keyframes` at-rule while collecting coverage (@scf2k).
-
Replace
GraphicsMagickwith lightweightpng-imglibrary (@j0tunn). As for this version,geminino longer requires any external tool for image processing. -
CSS coverage now supports source maps (@scf2k). If your CSS has the sourcemap, coverage report will show original files.
You can use
sourceRootoption to tellgeminiwhere are your sources located on a filesystem. -
CSS coverages statistics will now also be written to
coverage.jsonfile (@scf2k). -
Add
coverageExcludeoption to allow exclude certain files from coverage report (@scf2k). -
Add index page to coverage report and improve appearance of a the coverage pages (@scf2k).
-
Restructure documentation and add quick start guide in russian and english (@jk708).
- Reset cursor position before reloading the page. When resetting it just after page load, old cursor position may trigger some transitions which will not be finished before first screenshot.
testcommand will exit with correct codes
-
New image comparison algorithm is implemented:
gm comparereplaced with custom diff. For now, GraphicsMagick is still required for other image manipulations.tolerancesetting is removed in favor of "strict mode": by default, only noticable changes (according to ciede2000 metric) will be treated as failure, to treat all changes that way user can enablestrictComparisonoption.- in case if some element in focus during test can potentially have blinking caret displayed, diff will try to ignore caret.
-
Remove built-in TeamCity reporter. If you really want it, you can adapt 0.7.x reporter to the current version of
geminiand publish it in separate package. -
Remove legacy ability to specify browsers as array which was deprecated since 0.4.0.
-
Add experimental programmatic API.
-
Add
--grepoption togatherandtestcommands, which allows to execute only suites, matching the pattern (@scf2k). -
Show required dimensions in error message when origin does not fir to full-page screenshot (@scf2k)
- Show meaningful error when capture area origin does not fit to full-page screenshot. This change can break some of your tests (@scf2k).
- Add experimental CSS-coverage report. Enable
with
--coverageCLI flag orcoverage: truein config (@scf2k). - Add ability to override
gridUrl,rootUrl,screenshotsDir,debugconfig options by their respective CLI options or environment variables (@arikon). - Add config option
windowSizeto specify default size of the browser windows (@scf2k). - Add action
setWindowSizeto specify browser window size during the tests (@scf2k).
Was republished as 0.7.0 due to a breaking change.
- Add command line completion (@unlok).
- Show an error, when config file has unknown options.
- Add
focus(element)action (@arikon).
- Allow to use empty string with
sendKeys. It can be used to focus on an element without changing states.
- Reset mouse position for each suite. Previously, cursor may stay at the position left from a previous suite and some elements was captured with hover effect when there shouldn't be any.
- Fix incorrect capture region rounding, causing bottom row of the element to be cropped sometimes.
-
:beforeand:afterpseudo-elements outline and shadow are now taken into account when calculating capture region(@incorp).This change can break some of your tests. Re-gather reference images to fix the problem.
-
Added
toleranceconfig option which can be used to specify maximum error rate before images will be treated as unequal. Default tolerance to 0. This is stricter then previous versions: now every, even slightest difference between reference and current images will fail tests.Set
toleranceto 0.001 in.gemini.ymlto restore 0.5.x behavior. -
Coordinates of capture region are now rounded to capture maximal area. Previously, border pixels could be cropped, due to rounding error. This change can break some of your tests. Re-gather reference images to fix the problem.
-
Added ability to change diff highlight color in config file.
-
Change diff highlight style
-
Global installations of
gemininow runs local one, if available.
- Browsers are now launched once for each run (previously, they were
launched once per suite). This greatly reduces total tests run time,
but can break some of your code, i.e. each
mouseDownshould always be closed bymouseUp. Previously, this was not required ifmouseDownwas used once for suite. It will show warning if versions of the modules does not match. flatreporter replacestree. Tree reporter can not work with new browser launch model.- Add
parallelLimitoption that allows to limit number of browsers run in parallel. - Add
suite.after()which can be used to perform some action after all of the states without taking screenshot. - Unknown errors, returned by Selenium have more detailed report.
- Fix
--versionoption.
- Fix missing images in html report.
- Correctly detect crop region in Firefox
-
Crop region for screenshots is calculated via client script inside browser instead of
gemini. This allows to issue fewer requests to Selenium Server speeding up total tests run time. This feature breaks compatibility with old browsers (IE< 9). -
New config format, which allows to specify full set of capabilities for browsers:
browsers: phantomjs: phantomjs opera12: browserName: opera version: '12.06' platform: 'WINDOWS' firefox28: browserName: firefox version: '28.0' firefox27: browserName: firefox version: '27.0'
-
Correctly capture screenshots of regions out of initial browser viewport in browsers, that can't capture full page (
OperaandChromeat the time of writing). -
outline-widthof an elements is now also taken into account when calculating crop region. -
Add
debugoptions to config file. If set totrue,geminiwill print debug logs to STDOUT. (@arikon). -
Add
httpsection to config file which allows to configure HTTP timeout (http.timeout) retry count (http.retries) and delay between retries (http.retryDelay). (@arikon). -
Asynchronous errors stacktraces in browser actions (such as not found element) will point to users code.
-
More HTTP requests to Selenium will run in parallel speeding up
gather/testcommands (@arikon). -
If
geminiis run without subcommand, help text will be shown.
- Enhanced html report:
- suites are now collapsible;
- all but failed suites are collapsed by default;
- buttons to expand all, collapse all and expand only errors are added;
- stats of total numbers of tests run, failed, succeeded and skipped are shown at the top.
- Allow to use multiple reporters in
testcommand. - Throw error when creating multiple suites of the same name within the same parent.
- Throw error when creating multiple states of the same name within the suite.
- Throw error when creating suite that will be unable to run ( has states and hasn't url or capture region);
- Check argument types of
SuiteBuildermethods. - Check argument types of all actions methods.
- Shorter stacktraces for invalid elements errors.
- Correctly handle offsets in
mouseMoveactions. - Fix error when
gridUrlwas required even if there is onlyphantomjsbrowser.
- Allow to override
gridUrlandrootUrlsettings with cli options--grid-urland--root-url. - Correctly report error, when wrong argument passed to an action.
-
Ability to set additional capabilities for all browsers, using
capabilitiesoption in.gemini.yml:capabilities: option1: value, option2: value
-
Non-existent directories, passed to
gatherandtestcommands will be filtered out -
If fatal error occurs,
geminiwill always exit with 1 status code -
If test fails or state error occurs,
geminiwill always exit with 2 status code. -
When
geminiis unable to launch browser, more clearer error message will be displayed.
-
Elements to take screen shots of and elements to perform action on are now defined differently.
setElementsandsetDynmaicElementsmethods removed.New way to define elements for screenshot:
suite.setCaptureElements('.selector1', '.selector2', ...)
Or using array:
suite.setCaptureElements(['.selector1', '.selector2', ...])
To get element to perform action on, you can now pass selectors directly to the actions:
suite.capture('state', function(actions, find) { actions.click('.button'); });
To avoid multiple lookups for the same element you can use
findfunction which is now passed to the state callback:suite.capture('state', function(actions, find) { var button = find('.button'); actions.mouseMove(button); .click(button); });
-
Add
suite.before(function(action, find))which can be used to perform some actions before the first state. Context is shared between before hook and all of suite's state callbacks.You can use
beforeto look for element only once for the state:suite.before(function(actions, find) { this.button = find('.buttons'); }) .capture('hovered', function(actions, find) { actions.mouseMove(this.button); }) .capture('pressed', function(actions, find) { actions.mouseDown(this.button); });
Or to perform some actions before first state without taking screenshot:
suite.before(function(actions, find) { actions.click('.button'); });
-
Add
suite.skip()method which allows to skip suite in some set of browsers:suite.skip()- skip in all browsers.suite.skip('chrome')- skip in all versions of Chrome.suite.skip({name: 'chrome', version: '33.0'})- skip in Chrome 33.0suite.skip(['chrome', 'opera'])- skip in Chrome and Opera
-
Public API now has constants for special keys to use in
sendKeysactions (i.e.gemini.CONTROLforCTRLkey).
- Fix a bug with incorrect reference to the suite in states. Because of this bug dynamic elements was not updated properly.
-
New test suites API. Plans are replaced by test suites defined by explicit call.
Old version:
module.exports = function(plan) { plan.setName('some name') .setElements(...) .setDynamicElements(...) .capture(...) };
New API:
var gemini = require('gemini'); gemini.suite('some name', function(suite) { suite.setElements(...) .setDynamicElements(...) .capture(...) };
Suites also can be nested. In this case, child suite inherits all properties from the parent.
gemin.suite('parent', function(suite) { gemini.suite('child', function(child) { gemini.suite('grandchild', function(grandchild) { }); }); });
-
.reload()method is removed. Use nested suite if you need to reload browser. -
Added action to run any JS code in browser context:
actions.executeJS(function(window) { window.alert('Hello!'); });
-
sendKeysaction can optionally take an receive an element to send keys to.actions.sendKeys(elements.someInput, 'hello');
-
Added ability to specify browser version in
.gemini.ymlbrowsers: - {name: 'phantomjs', version: '1.9'}
-
Tree reporter is now used for
gathercommand.
phantomjsalways starts maximized. This fixes the error, when some shadows didn't fit in crop rectangle.- Action on dynamic element that does not currently exists causes non-fatal error. Such error will fail only one state, the rest will continue running.
gathercommand now reports browser name.- browsers are always closed, even if there was an error.
- Initial release