Version number notation guide:

The version number of the IEDriverServer.exe is in the pattern as follows:

major.minor.build.revision

The "major" and "minor" fields are synchronized to releases of the Selenium

project. Changes in the "build" field indicate interim releases made publicly

available via the project downloads page. Changes in "revision" field indicate

private releases checked into the prebuilts directory of the source tree, but

not made generally available on the downloads page.

v4.16.0.0

========

* Fixed typos in logged messages (#13120)

v4.14.0.0

========

* Fix potential null pointer access in CookieManager

v4.11.0.0

========

* Ignore process id match when finding the window handle - IE Mode on Edge. (#12246)

v4.8.1.0

========

* Update atoms (Use .textContent instead of .innerHTML in clear() action (#11504))

v4.8.0.0

========

* Remove unnecessary executable permission bit from several files (#11502)

v4.7.0.0

========

* Ignore Zoom Level for Edge IE Mode (#11055)

v4.6.0.0

========

* Enable msIEModeAlwaysWaitForUnload for Edge IE Mode (#11079)

v4.5.0.0

========

* Auto-locate Edge if IE mode requested and location not specified

* If system IE redirects to Edge, automatically use Edge in IE Mode

v4.3.0.0

========

* fix Edge IE mode issue that sometimes cannot get the handle of a newly opened window (#10702)

* Disable popup blocking for Edge in IE Mode (#10869)

v4.2.0.0

========

* Check process ID when searching for Edge window handle

v4.0.0.0

========

* (on behalf of Gavin Xu) Updated atoms to allow IE driver to find elements.

* (on behalf of Gavin Xu) Fixed use of NewWindow3 API for use with Microsoft

Edge in IE Mode.

* Made further improvements for using the driver with Microsoft Edge in IE

Mode.

v3.150.2.0

==========

* Updates to JavaScript automation atoms.

* Changed source to compile against Windows 10 SDK. This means that the

executable may no longer run on versions of Windows less than Windows 10.

v3.150.1.1

==========

* Made checking of the Content-Length header case-insensitive. This fixes

a number of issues where sessions cannot be started or where commands

fail because the HTTP client made a choice to not use properly-cased

header names.

v3.150.1.0

==========

* Reverted change to use effective style for determining element visibility.

If a containing element has scrollbars, simply using the client bounding

rect to determine the click point could lead the click point behind a

scrollbar be determined as "in view," when clicking on the point would

instead click on the scrollbar. A previous fix for this issue was not

sufficiently robust in all scenarios, leading some elements to be seen

as not displayed at all. Accordingly, this change has been reverted

until a more correct one can be created.

* Other updates to JavaScript automation atoms.

v3.150.0.0

==========

* (on behalf of Stanley Hon) Added basic support for IEDriver to drive

Chromium-based Edge in Internet Explorer mode. Users can do this by

setting the `ie.edgechromium` property to `true`, and by supplying

the path to the Chromium-based Edge executable in the `ie.edgepath`

property of the `se:ieOptions` capability object. Note that this

feature is only available on Windows, as Chromium-based Edge only

supports Internet Explorer Mode on Windows. Moreover, users cannot

switch session types from normal mode to Internet Explorer Mode or

vice-versa; only one session type is allowed.

* Made driver work with <map> and <area> elements

* (on behalf of Marc Fisher) Fixed FTP and HTTP proxy markers for setting

web proxy with the IE driver. Fixes issue #7152.

* Added additional error information to error messages encountered during

navigation.

* Removed support from IE driver for "desiredCapabilities" in new session

requests. The W3C WebDriver Specification does not recognize any property

in the payload for the new session command other than "capabilities". The

IE driver has long supported the presence of "desiredCapabilities" as a

fallback for older, obsolete language bindings. Now that we are in the

process of creating the Selenium 4.0 language bindings, this legacy

compatibility should no longer be necessary. Users using this version of

the IE driver with language bindings prior to 4.0 may encounter issues

if the bindings do not properly format the new session request;

nevertheless, this change will not be reverted, as future versions of the

language bindings should properly format the new session command.

v3.141.59.3

===========

* Made driver calculate viewport with window.innerHeight/Width instead of

using documentElement.clientHeight/Width. This follows the definitions of

the W3C WebDriver Specification for calculating the in-view center point of

an element.

* Fixed to return empty object from JavaScript instead of null.

* Fixed IE driver to ignore unknown timeouts for set timeout command.

According to the W3C WebDriver Specification, additional timeouts not

specified are valid syntax, though they are ignored by the driver.

* Updated driver to allow detection of alerts with 'do not create' checkbox.

v3.141.59.2

===========

* Slightly modified setting IE window to have focus when sending input.

This code only affects the browser when the requireWindowFocus

capability is set at session start.

v3.141.59.1

===========

* Added checking for max expiration date in IE driver cookie handling.

Attempting to add a cookie with a very large expiration date caused

the IE driver to crash because of overflows in the C++ standard library

time formatting functions. This commit now avoids the crash by

enforcing a maximum cookie expiration time of 2,147,483,647 (2^32 - 1)

seconds from the time the cookie is attempting to be set. At the time

of this writing (2019), that means an expiration of the cookie sometime

over 68 years into the future (2087). If the intended expiration time

is beyond that limit, the driver will now return "unable to set cookie"

error.

As an internal implementation detail, the driver is migrating from the

"expires" attribute to the "max-age" attribute of the cookie string,

as versions of IE below 9 are no longer supported. Additionally, while

it would be possible to use "expires" and get a longer expiration time,

(somewhere around the year 3000), the distinction between a cookie

expiring after 68 years and one expiring in just under 1000 years is

(or ought to be) largely meaningless in the context of a website.

Fixes issue #7122.

v3.141.59.0

===========

* Rollup of fixes since previous release. No additional changes.

v3.141.5.13

===========

* Fixed execution of user JavaScript to allow comment at end of line.

v3.141.5.12

===========

* (on behalf of Reinhold Degenfellner) Updated to allow retries for finding

the upload file selection dialog. Fixes issue #6991.

* Modified to prevent navigation to directories using file:// protocol.

Navigating to a directory will open a Windows Explorer window, not open

the directory in the IE frame, and there is no way to automate the

resulting window. This is a deliberate deviation from the W3C WebDriver

Specification, as it's impossible to yield spec-compliant behavior from

Internet Explorer.

* Made driver return lowercase utf-8 for charset in content-type header.

v3.141.5.11

===========

* (on behalf of Reinhold Degenfellner) Fixed crash when attempting to

upload files. Fixes issue #6976.

v3.141.5.10

===========

* (on behalf of Ben Kucera) Updated proxy settings to set proxy bypass

addresses. Fixes issue #6086.

* Updated to be able to close Windows 10 credentials dialog. When Microsoft

updated Windows 10's credentials prompt dialog, they made it a dialog

that is not a standard alert dialog and therefore not able to be

automated using standard Windows automation techniques. With a recent

preview release, it became possible to manipulate this dialog using

the UI Automation feature of Windows. The driver will now be able

to accept or dismiss this credentials dialog now. Please note that

there is still no supported solution for entering credentials, as

there is no support for that feature in the W3C WebDriver Specification.

v3.141.5.9

==========

* Enabled serialization of the FileList object from file upload elements

(the return value of the files property in JavaScript).

* Improved recognition of Protected Mode boundary crossing on navigation,

particularly with URLs beginning with "about:blank"

v3.141.5.8

==========

* Improved driver use with invalid Protected Mode settings. Currently,

when the user does not set the Protected Mode settings of the browser

and sends the capability to bypass the checks for those settings,

the driver will attempt to predict when a Protected Mode boundary

will be crossed, and set in motion a process to reattach itself to

the newly created browser. This process is far from perfect. It is

subject to really challenging race conditions that are truly impossible

to eliminate entirely, because of the architecture of the browser

itself. Nevertheless, even in its flawed state, this is still a better

outcome than it was previously for users.

Please note that the advice and support policy of the IE driver will

continue to be that the user must set the Protected Mode settings of

the browser properly before using the driver. Any "issues" that arise

by not having the settings set, but that disappear when the settings

are corrected, are not considered by the project to be valid issues.

This will include, but not be limited to, issues like abandoned

browser instances not being closed, use of multiple instances of the

driver where the wrong browser window is connected to and automated,

and issues where the driver appears to hang upon navigation to a new

page. If the problem disappears when the browser is properly

configured, any issue reports will be immediately closed with a note

to properly configure the browser and remove the capability.

The following situations should be at least partially mitigated by the

change:

* Navigation to a new page

* Clicking on a link (specifically an <a> tag) that will lead to a

navigation to a new page

* Clicking a link that opens a new window

Other cases, like navigating backward and forward through the browser

history, clicking an element that submits a form, and so on, may not

be handled. In those cases, issue reports will be summarily closed,

unless a specific pull request fixing the issue is also provided.

Additionally, use of things like proxies to capture traffic between the

browser and web server may miss some traffic because of the race

conditions inherent in the mechanism used to reattach to a newly

created browser. Again, these race conditions are unavoidable, and

issue reports that are based on them will be immediately closed with a

note indicating that the browser must have its settings properly set.

These strict guidelines are not intended to be harsh, and are not put

in place with the intent to avoid investigating and fixing issues;

rather, they must be enforced because the underlying architecture of

the browser makes them unavoidable.

v3.141.5.7

==========

* Fixed uninitialized variable in driver.

v3.141.5.6

==========

* Fixed proper keyboard event values for extended keys.

v3.141.5.5

==========

* Added hack for handling cases where document.designMode is on.

In IE, when document.designMode is turned on, the browser is put

into a state where anonymous JavaScript functions cannot be run.

Obviously, this breaks a huge portion of the driver, so the driver

will now force documents out of designMode when attempting to

execute script. There should be vanishingly few real-world cases

where this has an adverse effect, with most sites using the

contentEditable attribute instead.

v3.141.5.4

==========

* Added hack for detection of IE in full screen mode.

v3.141.5.3

==========

* Again updated Unicode character processing for composition. The

previous change in 3.141.5.1 for using composed Unicode character

sequences on send and decomposed on retrieve was disturbingly

naive, and entirely incorrect. This revision mostly reverts that

change, with the difference of checking for "single-character"

encoding sequences when sending. The IE driver will no longer

produce the same input for some key combinations as other

drivers (see some Hangul sequences for concrete examples). The

mitigating factor of this difference in behavior, however, is

that the driver should now produce identical input and output

between sending keys and reading the text via either getting an

element's text or its attributes. Hopefully. There are likely

to continue to be corner cases where use of Unicode characters,

does not entirely match up to expectations. This is particularly

so in languages where individual code points can be combined to

form other glyphs identical to other individual code points.

v3.141.5.2

==========

* Enabled create new window command to create tabs. This change

overcomes the limitation of only creating top-level browser

contexts in new top-level windows. The IE driver can now open

new tabs when using the create new window command, as in the

.NET language bindings:

driver.SwitchTo().NewWindow(WindowType.Tab);

Note carefully that this change is only valid when explicitly

using the create new window command. It does not affect clicking

on links that open new top-level browsing contexts; those will

continue to open in new top-level windows. Furthermore, it is

imperative to note that this feature will absolutely not work

when Protected Mode properties are misconfigured. In other words,

if your code uses the .NET InternetExplorerOptions

`IntroduceInstabilityByIgnoringProtectedModeSettings` property

(or its equivalent in any other language bindings), this feature

will not work. Issue reports submitted regarding this will be

summarily closed.

v3.141.5.1

==========

* Implemented create new window commmand. The W3C WebDriver

Specification recently added the ability to create a new top-level

browsing context, or what appears to the user to be a new tab or

window. This change implements that command. Though the command

payload contains a hint argument as to what type of new top-level

browsing context the user desires (tab or window), the driver

implementation is free to ignore this type hint if it does not

support the type being requested. Since the IE driver has never

supported, and still does not support, creation of new top-level

browsing contexts using a new tab, in accordance with the spec, the

driver will create the new context in a new window, even if a new

tab is requested. Please note this is not a bug; it is as designed,

and is still fully compliant with the specification with this

behavior.

* Fixed edge case for clicking element. In the case of a fixed element

where the top-left corner of the element is outside the view port,

the click point calculation was not taking the offset of the top-

left when looking for the in-view center of the element. This change

fixes that issue.

* Updated Unicode character processing for combining characters. In a

previous revision, the IE driver was modified to normalize Unicode

strings that used combining characters to compose a single glyph when

sending keystrokes. This commit implements the reverse of that

operation when reading text data, decomposing the glyph into its

combining characters if required. This is a potential destablizing

change for text sequences that use combining characters, and care

should be taken by users to understand the potential differences if

handling text using such character combinations causes unexpected

results.

* Corrected error statuses returned for switching frames. The IE

driver now returns the correct status as specified by the W3C

WebDriver Specification for error conditions encountered when

attempting to switch frames.

* Modified to allow null value for script timeout. The W3C WebDriver

Specification has contradictory language regarding the script

timeout. On the one hand, it specifies that the value of the script

timeout can be `null`, indicating an indefinite timeout. On the

other hand, it specifies that when setting or getting timeouts, the

value of any timeout must be an integer between 0 and 2^53 - 1.

Since geckodriver has made the assumption that the former condition

is the correct interpretation, and the maintainers of geckodriver

have added a test to the W3C Web Platform Tests for WebDriver that

expects this interpretation, the IE driver is being modified to

follow that decision. It is the opinion of the developers of the IE

driver that this is the incorrect interpretation, but there is no

recourse to have the geckodriver maintainers revisit their decision.

v3.141.5.0

==========

* Added logging of JavaScript errors when executing scripts.

* Fixed incorrect error response format for invalid commands in IE.

Previously, when an invalid command was sent to the IE driver, the

response was not completely valid for the W3C dialect of the wire

protocol. In particular, the distinction was lost between an unknown

URL and an invalid HTTP verb, and this code was not properly converted

when the driver was updated to implement the specification. This change

fixes that. Correct responses are now sent back for both invalid

command cases. Additionally, the driver no longer hand-codes JSON for

these responses, instead creating a valid Response object, and using

the common serialization mechanism for transmission. Finally, this

commit includes some minor code formatting cleanup, which does not

affect the functionality of the driver. Fixes issue #6828.

v3.141.0.6

==========

* Modified to allow the driver to handle supplementary Unicode characters

in sendKeys. The IE driver lost the ability to check for surrogate

pairs after implementing spec compliance in 3.5. The driver will still

convert the keys being sent using sendKeys from UTF-8 to UTF-16 (Windows

development default for wide strings), but it will now properly

check for supplementary characters by checking for surrogate pairs.

* Moved normalization of Unicode strings to InputManager. Rather than

normalizing every conversion from UTF-8, it is more correct to only

perform the normalization when using the Unicode strings for keyboard

input. Accordingly, the normalization code is being moved from

StringUtilities to the InputManager class.

v3.141.0.5

==========

* Updated element obscured algorithm to handle shadow DOM polyfills. Shadow

DOM for IE is problematic. Shadow DOM is only available in IE as a

polyfill. If an element is part of a Shadow DOM (using a polyfill), IE's

elementsFromPoint implementation will show the component elements, not

necessarily the Web Component root element itself. If the direct parent

of the Web Component host element is in this list, then it counts as a

direct descendent, and won't be obscured. Note that a potential

enhancement to this change is to walk the DOM tree up until an ancestor

is found that does not have a shadow root, but that enhancement is not

implemented at present.

* Added synchronization code. The IE driver can only handle one command

at a time. Commands from multithreaded client libraries should block

on the second thread's command until the first thread's executing command

is complete.

* Reverted earlier change to the obscured element algorithm. The previous

changes were incorrect. Any value other than 'none' for the pointerEvents

CSS property on a non-SVG element will result in the element receiving

pointer events.

* Removed isFocusable atom from sendKeys command. The isFocusable atom

does not provide the correct information as to whether an element is

focusable according to the W3C and WHATWG DOM specifications. This

change removes the use of that atom, and relies instead on the abililty

to set focus to the element before sending the keystrokes.

v3.141.0.4

==========

* Fixed to correctly set Content-length header for HTTP responses.

Not setting this header properly is a protocol violation, and strict

HTTP clients (like .NET's HttpWebRequest/HttpWebResponse) will reject

such an HTTP response for the 1.1 version of the HTTP protocol.

* Reduced JavaScript errors during findElement(s). In most cases, executing

JavaScript is an atomic action for the IE driver. However, in the off

chance that there is a page refresh during the execution of JavaScript,

converting the result of the script to a JSON object for serialization

might cause an error, in which case we should return the appropriate

error to the caller. One place, however, where this might be common is

executing findElement(s) across a page navigation. This change makes the

driver return the proper error in the post-processing of a script

execution. It also recognizes the findElement(s) case, ignoring

JavaScript errors in finding elements, and returning the proper structure

(an empty list or a "no such element" error) in its place. This lets

things like explicit waits retry the find, which may succeed once the

page navigation has completed and the DOM available for querying again.

* Tweaked obscured element algorithm. The algorithm used by the IE driver

to detect obscured elements checks the value of the pointer-events CSS

computed style property. This property can have valid values of "auto"

and "none" which affect whether to consider an element "obscured."

However, the property can also contain other values for SVG elements.

If the element in question is not an SVG element, then values other

than "none" and "auto" must be ignored, and the element considered not

to obscure the target element.

* Fixed compile warnings.

v3.141.0.3

==========

* Updated Civetweb dependency to version 1.11.

* Modified IE driver to no longer always send Connection: close header.

This is a potentially destablizing change, as the driver now should

correctly use keep-alive for HTTP requests and responses. While this

should not matter, the driver has not ever executed true with keep-

alive connections, so care should be used.

v3.141.0.2

==========

* Updating IE scroll-into-view algorithm for element click.

This update makes the scroll-into-view algorithm for element click

to be compliant with the W3C WebDriver Specification.

* Completed IE implementation of strictFileInteractability capability.

By default, the IE driver can interact with hidden <input type='file'>

elements. By setting this capability to false, it forces the driver

to make sure the file upload element is interactable like other

elements.

* Updates to JavaScript automation atoms.

v3.141.0.1

==========

* Updated error message return from unserializable JavaScript results. In

the case where a user JavaScript execution completes successfully, but

the resulting option is not serializable via JSON (because of cyclical

references or similar), the error message that the execution errored due

to the failed serialization was being suppressed.

* Added support for strictFileInteractability capability. The latest

editors' drafts of the W3C WebDriver Specification as a living document

have introduce the "strictFileInteractability" capability for handling

<input type='file'> elements. This change makes the driver aware of that

capability.

* Removed premature break statements in obscured element check. When

looping through the elements returned by elementsFromPoint, the driver

was calling a break statement early, preventing clear error messages for

which element is obscuring the element to be clicked.

v3.141.0.0

==========

* Release to synchronize with release of Selenium project.

* Updates to JavaScript automation atoms.

* Version number updated to continue with "digits of pi" theme.

v3.14.0.15

==========

* Fixed change in behavior recognizing closing of showModalDialog windows.

Fixes issue #6574.

v3.14.0.14

==========

* Fixed retrieval of CSS properties that return non-string values. Fixes

issue #6527.

v3.14.0.13

==========

* Fixed multiple file upload element detection.

v3.14.0.12

==========

* Allowed duration property of pause actions to be omitted.

* Cleaned up variant copying for getElementAttribute command.

* Expanded inline element detection for obscured elements. This commit

expands on the hack that was introduced in 3.14.0.11. It turns out

that <label> elements are not the only ones to suffer from issues

with elementsFromPoint. Rather, we now check for all types of inline

elements. This change also refactors the calculation of the hit-

testing point for calling elementsFromPoint in the first place, as

it only makes sense to call from within the context of the element's

document (omitting frames).

v3.14.0.11

==========

* Added hack to obscured element check for label elements. As per the

W3C WebDriver Specification, the IE driver uses elementsFromPoint to

determine whether an element being clicked on is obscured. However,

there is a bug in the IE implementation of elementsFromPoint that if

the element parent tree contains a <label> element, that element is

omitted from the list of elements returned by elementsFromPoint. This

change is to work around that bug. The solution implemented here is

to check if the top-most element as returned by elementFromPoint is

a direct descendent of the element we want to click on, and the

element we want to click on is a <label>, we assume the element is

not obscured. Note carefully that this is a hack of the highest

order, and there's every likelihood that some page, somewhere will

fail this check. Upon receiving an issue report with a complete

reproduction case that includes a page clearly demonstrating the

issue, we will then revisit this algorithm.

v3.14.0.10

==========

* Partially reverted obscured element check. The updated obscured check

introduced in 3.14.0.8 assumed that there was a bug in IE's implementation

of elementsFromPoint. This is not the case. The actual bug is that the

driver did not take into account elements found outside the DOM tree of

the element being interacted with should not be flagged as obscuring the

target element if its pointer-events CSS property is "none". This is

because while the element may technically be rendered above the target

element in the DOM, it does not interfere with attempts to interact with

elements below it, since pointer events are not triggered.

v3.14.0.9

=========

* Added ability to return text of JavaScript errors. Until this change,

if the IE driver encountered an error in a user's JavaScript, there was

no way to return the actual error text. This change fixes that. The text

of the JavaScript error will be returned so it can be used as part of

the exception thrown by the language bindings.

v3.14.0.8

=========

* Updated obscured element check. The W3C WebDriver Specification dictates

that checking for elements being obscured by other elements should rely

on calling elementsFromPoint. However, there is a bug in IE's

elementsFromPoint implementation. if two sibling elements have the same

z-index, the one specified second in the DOM order is usually on top.

The exception is if there is a child element that has a specified

z-index greater than the ancestor z-index. The IE elementsFromPoint

implementation ignores the child element case, so we have to manually

look for child elements that may be rendered on top of the element found

outside the element DOM tree of the element we are attempting to interact

with.

v3.14.0.7

=========

* Instituted limits on x and y offsets for actions.

* Corrected finding of extracted dll on startup.

v3.14.0.6

=========

* Updated to make JSON serialization of JavaScript objects more compliant.

When returning values from user-specified JavaScript, the presence of a

toJSON() property indicates that function is what should be used to

serialize the object. Additionally, for objects that are not recognized

as JavaScript objects via the properties of the IDispatch, and do not

have a toJSON property, the driver will now make a last-ditch attempt

to serialize the value for return by calling JSON.stringify().

v3.14.0.5

=========

* Updated to manually initialize cookie manager on new window open. When

a new window is requested by the browser, the driver creates one, but

does not set the window handle of the window to a proper value. This

is normally not an issue, but in the edge case where the user wants to

manipulate cookies in the new window, the cookie manager does not have

a window handle set to facilitate the transfer of cookie information

from the driver code to the browser. This change rectifies that problem.

v3.14.0.4

=========

* Updated to return 500 for timeouts instead of 408. Some HTTP clients

would retry when receiving a 408 status code. Switching to 500 resolves

that issue.

* Updated driver for the case of clicking element with children. The

case where one attempts to click on an element that must be scrolled into

view, and where the element has child elements, but the first node is a

text node was broken for IE. In this case, to guarantee the proper element

is clicked (and that an embedded child does not receive the click), the

driver would attempt to click on the text node. However, the coordinate

calculation when the element had to be scrolled into view was flawed. This

commit changes from using the boundingLeft and boundingTop properties of

IHTMLTextRangeMetrics to using offsetLeft and offsetTop in an attempt to

correct the calculation of the location of the text to be clicked on.

* Updated driver to return non-integer values for element rect.

* Made driver spec compliant for setting largest timeout value.

* Updated to properly find XML documents for element enabled and CSS values

* Refactored opening and closing of new browser windows. This simplifies,

streamlines, and makes more robust the driver code responsible for finding

new browser windows when opened by users' code, like clicking a link that

opens the target in a new browser window. It also simplifies the code that

executes when a browser window is closed. There is a chance this change

introduces a slight performance delay in the process of opening a new

browser window, but in the interest of making the driver more robust, it

is preferable to live with the delay and figure out a way to reduce or

eliminate it later.

v3.14.0.3

=========

* Updated to properly validate duration property of pause actions.

v3.14.0.2

=========

* Robustness change for invalid command payloads to comply with W3C Spec.

* Updated to truncate obscured element click error message to a reasonable

length.

* Handle null pointers for cookie values. Fixes issue #6261.

v3.14.0.1

=========

* Updated to properly set the "path" property on cookies when a document

is specified in the property.

v3.14.0.0

=========

* Release to synchronize with release of Selenium project.

* Modified to prevent attempt to return IDispatch that is not IHTMLElement.

When returning values from JavaScript, if a variant value is an IDispatch,

but does not implement IHTMLElement, and also is not recognized as a

JavaScript array or object, avoid returning it from the script. This is

also to prevent returning things like a window or document reference, as

those items contain recursive references that cannot be properly

serialized to JSON.

* (on behalf of Mike Bellew) Fixed null pointer exception when managed

element is not found. The command executor has a chance to return an error

code of ENOSUCHELEMENT when executing scripts. This leaves the element

wrapper object equal to null, which in turn ends up throwing an exception

on the next like of code when it tries to use the element. This would

cause a crash in IEDriver and it shuts down

v3.13.0.5

=========

* Modified to silently dismiss onbeforeupdate dialogs. The W3C WebDriver

Specification dictates that all onbeforeupdate dialogs are automatically

dismissed without notification. This commit brings the IE driver into

compliance with that part of the spec.

* Resetting frame focus to top-level frame on refresh. The W3C WebDriver

Specification dictates that issuing the Refresh command will reset the

focus to the top-level frame. This commit brings the IE driver into

compliance with that part of the spec.

v3.13.0.4

=========

* Made set window rect command spec-compliant for x and y coordinates

v3.13.0.3

=========

* Fixed crashing bug in obscured element detection.

* Modified to extract IE internal library to IEDriver.tmp first.

If the file IEDriver.tmp exists in the destination directory, then use

the result of the `GetTempFileName` API, which results in a file named

"IEDxxxxx.tmp" where "xxxxx" is a random number.

* Added version resources to internal IE driver library.

v3.13.0.2

=========

* Modified to prevent keyup actions if key is not already down for IE.

This change now checks that a specified key is down before sending any

input events when a keyUp action is specified. This prevents spurious

events detected by the browser, events which cannot happen with an actual

user.

* Refactored IE sendKeys command handler to handle multiple file uploads.

Until now, the IE driver has not handled the "multiple" attribute for

<input type="file"> elements. This construct is now supported, and it is

possible to specify multiple files for upload if the multiple attribute is

set on the <input> element. To do so, use a string containing full path

and file name to each file to be uploaded, separated by new line

characters (e.g.: "C:\path\to\file\one.txt\nC:\path\to\file\two.txt"). One

caveat to this approach is that when specifying multiple files, all files

must be located in the same directory. This is not a limitation imposed by

the driver; it is a limitation imposed by IE itself. It is a known

limitation, and cannot be fixed in the driver until and unless the browser

allows it.

* Added wait for navigation to IE accept and dismiss alert commands. When

accepting or dismissing an alert, particularly when it is raised during

the onload event, it is possible for the driver to not recognize that

navigation is still pending once the alert is handled. This change is an

attempt at better allowing the driver to recognize this situation and act

accordingly.

v3.13.0.1

=========

* Updated to automatically dismiss onBeforeUnload event dialogs. This is a

change from previous behavior, where such dialogs were handled via the

standard alert handling commands. The W3C WebDriver Specification demands

that onBeforeUnload dialogs be dismissed automatically, with no input

allowed for the user. This may be a breaking change for some users who are

currently relying on the alert-handling commands to handle these alerts;

however, not implementing this behavior makes the driver not compliant with

the specificiation. Users in that state will have no choice but to modify

their code to accomodate the new, spec-compliant behavior.

* Updated the reset action to be spec-compliant. This means that when the

"Release Actions" command is sent, that pressed keystrokes and mouse

buttons are reset to the "up" position in the correct order.

* Corrected "invalid session id" response to be spec-compliant

v3.13.0.0

=========

* Release to synchronize with release of Selenium project.

* Updates to JavaScript automation atoms.

v3.12.0.5

=========

* Updating to indicate elements with 'pointer-events' style of 'none' as

obscured, triggering "element click intercepted" error response. This is

in keeping with the W3C WebDriver Specification that demands this behavior.

v3.12.0.4

=========

* Updates to JavaScript automation atoms.

* Use elementFromPoint for IE driver in determining obscured elements. This

is in direct violation of the algorithm specified in the W3C WebDriver

Specification, but it fixes a class of IE-specific bugs that present

themselves when using elementsFromPoint. This is used to shortcut the use

of elementsFromPoint if it is not necessary.

* Added additional logging to IE driver for setting window focus.

* Updated IE driver to use SendInput for a single input at a time. The IE

driver uses the Windows `SendInput` API when a session is initiated with

the `requireWindowFocus` capability. This is a low-level user input

simulation API, and very closely mimics the use of a physical mouse and

keyboard. One of the guarantees of the API is that when one sends an array

of `INPUT` structures, then no input messages can be inserted into the

middle of the input stream. However, it appears that there are cases where

sending multiple input structures that IE does not process all of them.

This leads to truncated or missing keys when using the Selenium sendKeys

method. This change makes the driver use SendInput, but only passing a

single INPUT structure at a time. This is dangerous, in that it allows

other inputs to potentially be slipped into the input stream. This change

makes the driver prefer one form of instability (potentially unwanted

input events) over another (input events not processed by the browser).

Additionally, this change introduces an additional method for forcing the

IE window into focus before sending input. This additional method uses the

Windows UI Automation accessibility API to do this. The potential down

side to using this (apparently) more reliable mechanism is that it may

introduce a slight performance penalty, as the UI Automation API is

notoriously slow. In practice, this performance penalty is observed to be

on the order of milliseconds, so the trade-off here is deemed worth the

risk, and there is no way to disable this new check.

v3.12.0.3

=========

* Refactored IE waiting for page load on click. This commit restores some

previously lost functionality for waiting for a page to load on click of

an element. The race condition for detecting when a navigation is started

is mitigated by a short wait (50ms) on the click.

* Made "eager" page load strategy not wait for frame navigation to be

completed before returning.

v3.12.0.2

=========

* Added log warning for mismatched bitness of driver and browser when

sending keys.

* Updated to better support using CreateProcess API for launching IE.

The driver checks the registry for the location of the iexplore.exe

executable. In the case of a 32-bit driver on 64-bit Windows, the

registry calls are redirected to the Wow6432Node in the registry,

which ultimately points to ielowutl.exe instead of iexplore.exe. In

this case, we will now bypass the registry redirection, and read the

proper location from the 64-bit browser executable. A side effect of

this is that users wanting to use the CreateProcess API will need to

use the 64-bit version of the IE driver on 64-bit Windows, contrary to

most expectations. It is important to note that using this API and

launching the browser in this way is not consistent with how the user

will launch IE, and if the driver is used for testing purposes, it will

not be giving identical results to what the end-user of the application

being tested will use when running IE.

v3.12.0.1

=========

* Update close window command to return list of remaining window handles.

This makes the command consistent with the W3C WebDriver Specification.

Unfortunately, to accurately return the proper list, however, this

requires the driver to wait for the window to be completely closed. In

previous versions, this was an asynchronous operation, and could return

immediately. With this change, the driver must wait for the window to

be successfully closed, and this change therefore introduces a slight

performance delay when windows are closed. The performance hit is

unfortunate, but unavoidable.

v3.12.0.0

=========

* Release to synchronize with release of Selenium project.

* Updates to JavaScript automation atoms.

* Added alert text to error response for unexpected alerts. This is to

maintain consistency with other implementations.

* Added checking of variant type for returns from JavaScript execution.

In order to prevent false positives when checking whether a returned

variant is an array or object, make sure the returned type of the

variant is VT_DISPATCH.

v3.11.1.6

=========

* Changed to throw an error for key actions taking multiple characters in IE.

The key actions as defined by the W3C WebDriver Specification take a JSON

payload that requires the resulting key for the key down or key up action

to resolve to a single Unicode code point or a single grapheme cluster

(what appears to be a single logical "character" from a user's

perspective). The IE driver will now evaluate the item passed into the key

down or key up action, and, taking combining characters into account,

attempts to make sure there is only a single grapheme present, if more

than one Unicode code point is sent. If the entered string would results

in multiple grapheme clusters, the driver now returns an invalid argument

error, as per the spec.

* Updated find element by link text atom to handle non-breaking spaces.

The WebDriver W3C Specification states that when finding by link text, the

text of a link element (<a>) should be trimmed of leading and trailing

whitespace after calling getVisibleText on the element before comparison.

Links with text that contains leading or trailing non-breaking spaces

(&nbsp) were having the non-breaking spaces converted to regular spaces,

but not trimmed after the fact. This commit corrects that oversight.

* Made IE cookie expiration times integer values. Some WebDriver

implementations have used floating point values for the expiration time of

a cookie as serialized in the wire protocol. The W3C WebDriver Specification

dictates that expriation times should be integer values. This commit makes

the driver conform to the specification.

* Modified to throw exception clicking on file upload elements in IE.

The WebDriver W3C Specification states that if a user attempts to click a

file upload element (<input type="file">), an error should be returned.

This is consistent with the intended behavior of WebDriver from the start,

which is that to upload a file, one should use sendKeys, not click() on

the input element. This now codifies that behavior with returning an

error. Note that if users have erroneously been using click() with file

upload elements, then using some external library (Robot Framework,

Sikuli, etc.) to handle the resulting file selection dialog, their

code will now break, in that the click() method will throw an exception,

and the element will not be clicked. Those users should update their code

accordingly.

v3.11.1.5

=========

* Initial implementation of taking element screenshots. This is an extremely

naive implmementation, and should not be considered complete.

v3.11.1.4

=========

* Fixed finding child elements to return proper error if parent element is

stale. Fixes issue #5700.

v3.11.1.3

=========

* Fixed detection of obscured elements when top element is not displayed.

Fixes issue #5668.

v3.11.1.2

=========

* Changed COM variant to JSON serialization in IE. There is an expectation

that if a result from JavaScript is "integral," that it is serialized for

return to the calling code as an integral number rather than a floating-

point number. Previous versions of the JSON serializer used by the IE

driver would (incorrectly) serialize a floating-point value with no

decimal part as being an integer. After updating to a later version, the

serializer now (correctly) serializes floating-point values as floating-

point values, with a trailing ".0" to indicate a floating-point value.

While this is the correct behavior by the serializer, it breaks the

expectations of the language bindings. Therefore, we check to see if the

variant value has a decimal part, and if it has none, convert it to an

integer for serialization by the JSON library.

* Updated JsonCpp library to latest source code. Fixes issue #5664.

v3.11.1.1

=========

* Separated out detection of focusable elements for sendKeys in IE. Allows

separate error messages for attempting to send keys to a non-focusable

element. Also fixes sending keystrokes to <option> elements by forwarding

the detections to the parent <select> element.

* Fixed obscured element detection in IE for elements in frames.

* Added support for extension capabilites starting with "test:" in IE. This

is basically to support the web-platform-tests WebDriver test suite. Adding

an extension capability with name starting with "test:" will allow a user

to create a session, but will have no effect on the session. It is a no-op.

* Added IsFocusable method to IE driver Element class.

* Added Protected Mode boundary crossing detection mechanism in IE. When a

Protected Mode boundary is crossed by navigation (entering or exiting

Protected Mode), the existing browser instance is destroyed and a new one

created in its place by Interent Explorer. This commit adds detection for

when a browser instance is being closed, but without an explicit call to

the WebDriver close() or quit() methods. When this is detected, it's likely

a Protected Mode boundary is being crossed, and all subsequent commands in

the WebDriver session will fail. In this case, the driver will now write

to the log that the browser has been asked to exit without the user

explicitly requesting it. The detection is not perfect, since it's

possible to legitimately click a link that closes the browser window, and

this is indistinguishable from clicking a link that navigates to a URL

that causes a Protected Mode boundary crossing. Nevertheless, this logging

is being added so that users can see what may be happening when they

receive errors like "Unable to get current browser."

v3.11.1.0

=========

* Updated all C++ projects for the IE driver to use the Windows 8.1 SDK.

This has the unfortunate side effect that the driver executable will no

longer launch on Windows XP.

* Removed direct call to InternetGetCookieEx2 API. Versions of IE prior to

11 do not have access to this API. Linking directly to it via the Windows

8.1 SDK creates an import entry in the IEDriver.dll library used by the

driver. So that the driver can continue to be used against earlier releases,

we will call the API using GetModuleHandle and GetProcAddress only if it

exists in the version of WinINet being used by the browser. Fixes issue #5603.

* Properly initializing JSON values in IE driver. Fixes issue #5519.

v3.11.0.0

=========

* Release to synchronize with release of Selenium project.

* Added additional logging to new session command.

* Added detection of intercepted element clicks to IE. This brings the IE

driver into alignment with the Chrome and Firefox (geckodriver) drivers,

in that it detects when an element attempting to be clicked on is covered

in the z-order by another element. Do note that this may cause different

behavior from previous versions of the IE driver for WebDriver code that

runs only on IE. For code that runs cross-browser, this should now yield

the same behavior across platforms.

v3.10.0.2

=========

* Updates to JavaScript automation atoms.

* Fixed clear element command to handle cases where the element's parent

document is undefined.

* Fixed sending keystrokes with native events disabled. Because apparently,

the driver's developer can't remember how function binding works in C++

when using a pointer to the object instead of the object itself. Fixes

issue #5584.

v3.10.0.1

=========

* Updates to JavaScript automation atoms.

* Updated clear element command handler to return W3C compliant errors for IE.

* Updated IE error response to be spec compliant for sendKeys

* Fixed IE cookie retrieval for IE 10 and 11. Internet Explorer relies on a

Windows component called WinINet to communicate over the Internet. This

includes WinINet APIs for getting and setting cookies. Prior to Windows 10,

persistent cookies were stored in files on the hard drive. A change in

WinINet moved the persistent cookie storage from files on the disk to an

Extensible Storage Engine (ESE or "Jet Blue") database, which rendered them

unreadable by the IE driver. This commit restores the functionality of

getting cookies by using an API function called InternetGetCookieEx2.

This API function was introduced in the Windows 8.1 SDK, which means that

it won't be available to older versions of WinINet. In that case, we fall

back to the original behavior of attempting to parse cookie files from the

disk. There is a chance that the detection of the availability of this API

is flawed in the IE driver.

* Revamped handling of modifier keys. The IE driver is now able to

distinguish between the left and right modifier keys (Shift, Control,

Alt) on the keyboard.

* Modified mouse event handling to match W3C Specification.

THIS IS A POTENTIAL BREAKING CHANGE.

The simulation of mouse movement for the actions command now conforms

to the W3C WebDriver Specification for mouse movement. This means

that offsets in elements are now measured from the center of the

element instead of the top-left, which was the previous behavior.

Additionally, attempting to move the mouse pointer outside the

browser view port will result in a "mouse movement out of bounds"

error. While users should be aware that this might cause code

using user interactions to fail, not that the behavior is now

consistent with the behavior outlined in the specification, and

the behavior of geckodriver in particular.

* Modified processing of pause actions in keyboard action sequences.

The vast majority of use of pause actions in a keyboard action

sequence are used to synchronize the entire action chain with other

(mouse) action sequences. To allow the driver to properly synchronize

the input sequences, the driver will now ignore the duration property

of pause events in keyboard action sequences. This is deliberately in

violation of the W3C specification, and may be revisited in the future.

* Fixed sending keystrokes outside the Basic Multilingual Plane. This

allows emoji and other Unicode characters that are represented by

surrogate pairs in UTF-16 to be sent to IE.

* Fixed getElementProperty to return proper data type. Previously,

the getElementProperty command would coerce the property value to a

string. This is not correct behavior as specified in the W3C spec.