Hi,
selenium-shutterbug v1.6, used with Firefox as part of a Selenium test suite. Selenium 4.36.0, in case that's relevant. We've got a wrapper that takes screenshots at various stages of test execution, and also when a test failure occurs.
We've just hit an issue where a flaky test, lacking a timeout, is reporting org.openqa.selenium.JavascriptException: TypeError: can't access property "scrollWidth", document.body is null
Here's the relevant stacktrace:
java.lang.NullPointerException
at com.assertthat.selenium_shutterbug.utils.web.Browser.getDocWidth(Browser.java:799)
at com.assertthat.selenium_shutterbug.utils.web.Browser.takeFullPageScreenshotScroll(Browser.java:206)
at com.assertthat.selenium_shutterbug.core.Shutterbug.shootPage(Shutterbug.java:163)
at com.assertthat.selenium_shutterbug.core.Shutterbug.shootPage(Shutterbug.java:83)
at com.assertthat.selenium_shutterbug.core.Shutterbug.shootPage(Shutterbug.java:69)
That's from a call Shutterbug.shootPage(driver, Capture.FULL_SCROLL).withName(filenameBase).save(outputFolder.getAbsolutePath());
Obviously, the root cause here is the missing timeout in our test, resulting in trying to take a screenshot at an inappropriate time, but it'd be great if shutterbug were tolerant of being asked to take a screenshot of a non-existent page.
I tried adding a check that document.body != null before calling shootPage, but that failed when the browser state changed after the check.
It's difficult to create a minimal example that reproduces this (but I am trying) and I realise that there are probably many code paths that can result in similar errors.
For this particular case, though, I wonder whether it'd be sufficient to make the JS paranoid so that max-document-width.js does something like Math.max(document.body?.scrollWidth ?? 0, document.body?.offsetWidth ?? 0, ...etc...), and similar for max-document-height.js, and then throw an UnableTakeSnapshotException if docWidth == 0 || docHeight == 0.
Hi,
selenium-shutterbug v1.6, used with Firefox as part of a Selenium test suite. Selenium 4.36.0, in case that's relevant. We've got a wrapper that takes screenshots at various stages of test execution, and also when a test failure occurs.
We've just hit an issue where a flaky test, lacking a timeout, is reporting
org.openqa.selenium.JavascriptException: TypeError: can't access property "scrollWidth", document.body is nullHere's the relevant stacktrace:
That's from a call
Shutterbug.shootPage(driver, Capture.FULL_SCROLL).withName(filenameBase).save(outputFolder.getAbsolutePath());Obviously, the root cause here is the missing timeout in our test, resulting in trying to take a screenshot at an inappropriate time, but it'd be great if shutterbug were tolerant of being asked to take a screenshot of a non-existent page.
I tried adding a check that
document.body != nullbefore callingshootPage, but that failed when the browser state changed after the check.It's difficult to create a minimal example that reproduces this (but I am trying) and I realise that there are probably many code paths that can result in similar errors.
For this particular case, though, I wonder whether it'd be sufficient to make the JS paranoid so that max-document-width.js does something like
Math.max(document.body?.scrollWidth ?? 0, document.body?.offsetWidth ?? 0, ...etc...), and similar for max-document-height.js, and then throw anUnableTakeSnapshotExceptionif docWidth == 0 || docHeight == 0.