You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
[Bug]: Integration tests with copyImage timeout in Firefox in headless mode #20918
When using Puppeteer with Firefox in headless mode, navigator.clipboard.write([new ClipboardItem({ "image/png": blob })]) succeeds without error, but subsequent keyboard paste (Ctrl/Cmd+V) fails to insert the image. The same code works correctly in Chrome (headless or not) and in Firefox non-headless mode.
We should either disable these tests for Firefox in the headless mode or use #stampEditorFileInput to add the image instead, except when testing the copy/paste functionality itself.
I ran into this in the initial version of #20371 too (which is #20029), and found out that this is bug in Firefox specifically. It's tracked in https://bugzilla.mozilla.org/show_bug.cgi?id=1878643 upstream, and eventually decided that for the short term only running the integration tests in headful mode is feasible because this might take some time to get fixed. I believe it's also not the only problem in headless mode, but it is the most prominent one at least, and it would be nice to strive for being able to run the integration tests successfully in both modes because ideally there should not be any difference.
Closing since the upstream bug is fixed. I have confirmed that most stamp editor integration tests, that invoke copyImage, now pass in Firefox headless mode. The other issues that prevent us from actually using headless mode in general for the integration tests are now tracked in #21991.
Activity
timvandermeij commentedon Mar 19, 2026
I ran into this in the initial version of #20371 too (which is #20029), and found out that this is bug in Firefox specifically. It's tracked in https://bugzilla.mozilla.org/show_bug.cgi?id=1878643 upstream, and eventually decided that for the short term only running the integration tests in headful mode is feasible because this might take some time to get fixed. I believe it's also not the only problem in headless mode, but it is the most prominent one at least, and it would be nice to strive for being able to run the integration tests successfully in both modes because ideally there should not be any difference.
timvandermeij commentedon Sep 20, 2026
Closing since the upstream bug is fixed. I have confirmed that most stamp editor integration tests, that invoke
copyImage, now pass in Firefox headless mode. The other issues that prevent us from actually using headless mode in general for the integration tests are now tracked in #21991.