User Agent: Mozilla/5.0 (Windows NT 6.1; WOW64; rv:25.0) Gecko/20100101 Firefox/25.0 (Beta/Release) Build ID: 20131112160018 Steps to reproduce: Moving a child element for example -10px using transform: translate() and at the same time moving the parent element 10px also using transform: translate(). Here is an example: http://jsbin.com/IsIVoyox/1/edit (hover the red elements) Bug appears in Firefox 25.0.1 and Firefox 28.0a1. Works as it should in Chrome, Safari and Internet Explorer 10. Actual results: When the child element goes outside of the parent element it will disappear. This however does seem to depend on the size of the child element and the distance it's moved. Expected results: The child element should still be visible when outside the parent element.

Confirmed on Mac too.

Works fine with 2013-02-01 nightly. So I think the next step is to get a regression range here.

Works in Firefox 21.0b7 but not in 22.0 Tested on win7

Great. If we could get that bisected down to a single day of changesets that would be amazing. mozregression can help do it quicker, instructions here http://mozilla.github.io/mozregression/

Last good revision: eccf45749400 (2013-03-11) First bad revision: 7433bc4545c9 (2013-03-12)

Thank you! That gives these changesets http://hg.mozilla.org/mozilla-central/pushloghtml?fromchange=eccf45749400&tochange=7433bc4545c9

Hmm, the only changeset in there that looks like it could have affected this is http://hg.mozilla.org/mozilla-central/rev/9a9689e414e4 form bug 849263.

...which was making a previous change that Matt made actually do what it was supposed to do.

(Though I'm actually a bit skeptical of that being the cause, since it really should have changed perf rather than behavior.)

There were actually two regression ranges, afaict. First one is indeed http://hg.mozilla.org/mozilla-central/pushloghtml?fromchange=eccf45749400&tochange=7433bc4545c9 where the behavior changed to the the blue square disappearing while the red one is moving but reappearing once the red one stops. Second one, to the current behavior: http://hg.mozilla.org/mozilla-central/pushloghtml?fromchange=97cfc16ba5dc&tochange=c232bec6974d

Local bisect says that either bug 849263 or the landing/backout of bug 804944 that happened at the same time triggered the first regression. My money is on the former.

Fixed by bug 944291.

Verified as fixed on latest Aurora 28.0a2 as well using Ubuntu 12.04 x32 and Windown 8.1 x32