I've been running into a lot of intermittent failures when loading pages in Chrome. The Page.ToUrl and Page.Wait methods appear to return almost immediately, especially when reloading a URL that is already being displayed in the browser. There is an additional problem in our application in that in-page scripting is used to handle links/buttons, so there is a slight delay before a page action causes the resulting page to load. The Page.Wait method would return before the page actually began loading, so it was not waiting for the new page to appear.To work around this, I tried waiting for the page to start loading by checking for the contentDocument.readyState property of the Page object to become anything other than "complete" in a script loop. This, in turn, led to a lot of script exceptions, such as "the object does not exist", "the object has disconnected from its clients", and "the object does not support this property". It appears that the contentDocument object is destroyed during the page load.I then tried using a script loop to check when the Page.contentDocument property became invalid (Nothing), but this lead to "Unspecified error" exceptions in the script.Can anyone think of any other options other than inserting a hard-coded delay before calling Page.Wait? Also, if anyone else is having similar issues with Chrome, it might be useful to chime in.Thanks,--Rachel

Hi Rachel,I looked at your code and my current understanding (without seeing what is going on with the actual page) is that it is worth reading the thread I recommended.In a nutshell: both yours and my code calls WaitPage() method. This method waits for the indication from the browser that it (browser) received the requested page. This means, that after the call to WaitPage the requested page (i.e. .html, .asp, .jsp, or whatever else one) is already received by the browser. But the fact that the page was received by the browser does not mean that the page is displayed to the end-user. In reality, especially with those modern 'fancy' and (over-) scripted applications, the browser starts to produce the content to present and it is possible that it meets some script code that does another request for the data to be included in the page. Note, that this another request has no relation to the page that was requested originally, i.e. the original page is already received and the fact that some script requests some data to be added to the page does not change original page's status (which is already 'ready').The browser continues to process the script code (and to execute additional requests) until all code is processed and only then the page is displayed to the end user.But this is not the end of the story, as the browser might receive data asynchronously and modify the contents of the page correspondingly. Note, that these modifications do not change original page's status either.Summarizing: if you are unlucky (and, most probably, you are like any other tester of the modern web application), then, in your test code, you must wait not only for the page to be received, but you also must explicitly search for and wait for the reasonable amount of time for the object on the page that you need to work with, as this object may appear not when the page is received from the server, but some time later, when the browser executes additional requests and processes additional data.Hope, the above will help...

I tried the second method out today, and it also seems to work across browsers. Get a top-level panel, loop and wait for it to stop existing, then call Page.Wait to wait for the browser page load to finish.I'll attach the new version as well. I think it's a bit cleaner, as it doesn't require a bunch of script exception handling.

Hi Rachel, Does the issue occur in TestComplete 9.31? Your page may contain many complex objects that are loaded for a long time. Try using the approaches to obtain objects specified in the "Testing Dynamic Web Pages" article.

Hi, Tanya,Our test platform was running TestExecute 9.30, so I have updated it to 9.31. This solved the issues with Page.ToUrl, so that's good.This leaves my problem of waiting for a page to load after clicking a script-based link. Example: There is a "Log out" link on application pages that ends the user's session and redirects them to the home page. If I click the link and immediately call Page.Wait, the method sometimes returns immediately, before the page redirect. My object searches then find the objects on the old page and the test fails.I've tried several different methods of determining when the page has started loading before calling Page.Wait, and they have all failed in Chrome 30. They all work in IE10 and Firefox 24.- Loop to check that Page.contentDocument.readyState is anything but "complete".Failed with "object does not exist" or "object does not support property" because the contentDocument was being destroyed between the moment I checked whether it was valid and the moment I actually tried to check the readyState.- Loop to check whether Page.contentDocument is Nothing.Failed with random "Unspecified error" script exceptions. When I added native error handling to catch these exceptions, I had a random error occur that the Page object did not exist.Next step is to reorganize the loop to obtain a new Page object every time...

After a good deal of experimenting today, it looks like IE10 and Firefox 24 both work if I click a link and call Page.Wait, but Chrome consistently returns from Page.Wait too fast, while the page redirect is still occurring. I've come up with a workaround method that I call after clicking any script-based link in our application. It seems to handle most of the random exceptions I'd been encountering, although only extended use will prove it out. I'm not certain if this is a TestComplete/Chrome issue or if it's just a quirk of our web application. I'll attach the method here in case anyone else runs into a similar problem.

Errors loading pages in Chrome | SmartBear Community

11 Replies

AlexKaras
Champion Level 2
12 years ago
Hi Rachel,

I looked at your code and my current understanding (without seeing what is going on with the actual page) is that it is worth reading the thread I recommended.

In a nutshell: both yours and my code calls WaitPage() method. This method waits for the indication from the browser that it (browser) received the requested page. This means, that after the call to WaitPage the requested page (i.e. .html, .asp, .jsp, or whatever else one) is already received by the browser. But the fact that the page was received by the browser does not mean that the page is displayed to the end-user. In reality, especially with those modern 'fancy' and (over-) scripted applications, the browser starts to produce the content to present and it is possible that it meets some script code that does another request for the data to be included in the page. Note, that this another request has no relation to the page that was requested originally, i.e. the original page is already received and the fact that some script requests some data to be added to the page does not change original page's status (which is already 'ready').

The browser continues to process the script code (and to execute additional requests) until all code is processed and only then the page is displayed to the end user.

But this is not the end of the story, as the browser might receive data asynchronously and modify the contents of the page correspondingly. Note, that these modifications do not change original page's status either.

Summarizing: if you are unlucky (and, most probably, you are like any other tester of the modern web application), then, in your test code, you must wait not only for the page to be received, but you also must explicitly search for and wait for the reasonable amount of time for the object on the page that you need to work with, as this object may appear not when the page is received from the server, but some time later, when the browser executes additional requests and processes additional data.

Hope, the above will help...
rgratis
Frequent Contributor
12 years ago
I tried the second method out today, and it also seems to work across browsers. Get a top-level panel, loop and wait for it to stop existing, then call Page.Wait to wait for the browser page load to finish.

I'll attach the new version as well. I think it's a bit cleaner, as it doesn't require a bunch of script exception handling.
waitforload2.zip1 KB
TanyaYatskovska
Alumni
12 years ago

Hi Rachel,

Does the issue occur in TestComplete 9.31?

Your page may contain many complex objects that are loaded for a long time. Try using the approaches to obtain objects specified in the "Testing Dynamic Web Pages" article.
rgratis
Frequent Contributor
12 years ago
Hi, Tanya,

Our test platform was running TestExecute 9.30, so I have updated it to 9.31. This solved the issues with Page.ToUrl, so that's good.

This leaves my problem of waiting for a page to load after clicking a script-based link. Example: There is a "Log out" link on application pages that ends the user's session and redirects them to the home page. If I click the link and immediately call Page.Wait, the method sometimes returns immediately, before the page redirect. My object searches then find the objects on the old page and the test fails.

I've tried several different methods of determining when the page has started loading before calling Page.Wait, and they have all failed in Chrome 30. They all work in IE10 and Firefox 24.

- Loop to check that Page.contentDocument.readyState is anything but "complete".

Failed with "object does not exist" or "object does not support property" because the contentDocument was being destroyed between the moment I checked whether it was valid and the moment I actually tried to check the readyState.

- Loop to check whether Page.contentDocument is Nothing.

Failed with random "Unspecified error" script exceptions. When I added native error handling to catch these exceptions, I had a random error occur that the Page object did not exist.

Next step is to reorganize the loop to obtain a new Page object every time...
rgratis
Frequent Contributor
12 years ago

After a good deal of experimenting today, it looks like IE10 and Firefox 24 both work if I click a link and call Page.Wait, but Chrome consistently returns from Page.Wait too fast, while the page redirect is still occurring. I've come up with a workaround method that I call after clicking any script-based link in our application. It seems to handle most of the random exceptions I'd been encountering, although only extended use will prove it out.

I'm not certain if this is a TestComplete/Chrome issue or if it's just a quirk of our web application. I'll attach the method here in case anyone else runs into a similar problem.

waitforload.zip1 KB
TanyaYatskovska
Alumni
12 years ago

Hi,

I've googled this - it looks like the issue is that the readyState property always returns complete in Chrome. That's why, your script is executed so fast.

So, I think there is nothing TestComplete can do here.

You may consider using another approach to wait until the page is loaded, for example Page.Wait.
AlexKaras
Champion Level 2
12 years ago
Hi Rachel,

You may consider this thread as an additional possible approach: http://smartbear.com/forums/?forumid=81&threadid=83955#83984.

As for the value of the readyState property: it is my understanding, that the browser changes the type of the parent contentDocument property while it renders the initial contents of the page and processes scripts on the page.

After some considerations, I ended up with the following code snippet that works stable so far for me:

If ("iexplore" = Aliases.browser.ObjectIdentifier) Then

' DONE: Check PageObject.contentDocument.readyState <> "complete" for non-IE browser(s)

    PageObject.Wait

    While 4 <> PageObject.readyState

'        Set w = PageObject.Parent.WaitWindow("#32770", "*", -1, 500)

'        if (w.Exists) then _

'          w.Close

      Call aqUtils.Delay(500, "Waiting for the page to load from the web server...")

    Wend

Else

    Do Until aqObject.IsSupported(PageObject, "contentDocument")

      Call aqUtils.Delay(500, "Waiting for the page to load from the web server...")

    Loop

    Do Until BuiltIn.varDispatch = aqObject.GetVarType(PageObject.contentDocument)

      Call aqUtils.Delay(500, "Waiting for the page to load from the web server...")

    Loop

    Do Until aqObject.IsSupported(PageObject.contentDocument, "readyState")

      Call aqUtils.Delay(500, "Waiting for the page to load from the web server...")

    Loop

    Do Until BuiltIn.varOleStr = aqObject.GetVarType(PageObject.contentDocument.readyState)

      Call aqUtils.Delay(500, "Waiting for the page to load from the web server...")

    Loop

    Do Until "complete" = PageObject.contentDocument.readyState

'        Set w = PageObject.Parent.WaitWindow("#32770", "*", -1, 500)

'        if (w.Exists) then _

'          w.Close

      Call aqUtils.Delay(500, "Waiting for the page to load from the web server...")

    Loop

End If

Hope it might appear useful for you...
rgratis
Frequent Contributor
12 years ago
I gave this a try, but it has the same effect as Page.Wait, which is that it completes before the page even starts loading in our application (in Chrome only).

Desired flow is:

- click javascript link

- wait for load of target page to be triggered

- wait for browser to finish loading new page

The function I attached is the best way I have found to wait for a page load to begin in Chrome before calling Page.Wait to wait for the browser to finish loading the page. If you have another method of waiting for a page load to start, please let me know.

Thanks,

--Rachel
rgratis
Frequent Contributor
12 years ago
I am familiar with attempting to wait for certain objects to become available, but this approach would require me to write separate code to verify every page transfer because the unique objects available would differ based on the page. Searching for common application page objects has the same problem as Page.Wait; WaitChild/FindChild sometimes return the object from the old page, before the new content has loaded.

I was really looking for a global function that would wait for a page load. The code I came up with is a bit of a hack, but it fixes the issues with Chrome 30+ and the Page.Wait function.

Another approach could be to get a reference to any top-level panel object on the original page, then wait for it to not exist (meaning the page load has occurred). I might explore this today if I have some spare time, as it would be preferable to messing around with Chrome's contentDocument and not require separate handling based upon browser.
safrrhmn
Occasional Contributor
12 years ago
try the following:

Sys.Browser.Page("*").Wait