With user_pref("layout.css.flexbox.enabled", true); Assertion failure: childDesiredSize.height >= aItem.GetBorderPaddingSizeInAxis(aAxisTracker.GetCrossAxis()) (Child should ask for at least enough space for border/padding), at layout/generic/nsFlexContainerFrame.cpp:1890

This drifted off my radar, probably because it uses huge nscoord math and hence is unlikely to be triggered by real content. I just inadvertently made a different (non-huge) testcase that triggers this same assertion, though, and I rediscovered this bug when checking for dupes. I'm going to file a different bug for the testcase that I made, since I'm not yet sure whether its fix will also address this bug. But while I'm here, I'm updating the original testcase for this bug to remove the "-moz" prefix, since "-moz-flex" isn't recognized anymore. (we unprefixed in bug 801098)

(In reply to Daniel Holbert [:dholbert] from comment #1) > I just inadvertently made a different (non-huge) testcase that triggers this > same assertion, though, and I rediscovered this bug when checking for dupes. > > I'm going to file a different bug for the testcase that I made, since I'm > not yet sure whether its fix will also address this bug. Filed bug 854263.

Note: Bug 854263 is reducing the severity of this situation to NS_ASSERTION, so we'll be able to proceed w/ reasonable behavior after hitting a situation like this. (This assertion continues to be possible to trigger with absurdly-large nscoord arithmetic, so it shouldn't really be fatal, but it does indicate something weird is going on so it's worth asserting about.)

Once bug 854263 downgrades the fatal assertion to a warning, I guess we'll consider this bug FIXED?

Sure, at least from the perspective of "this testcase no longer triggers assertion-failures". :) The "real" fix will be when we've got that nscoord-overflow flag on the document, and then we can conditionally disable this assertion when appropriate, as discussed on bug 854263.

I landed this bug's testcase as a crashtest, in a push with bug 854263's fix (which essentially fixes this, per prev. comment): https://hg.mozilla.org/integration/mozilla-inbound/rev/14064d4b31e7

(We can close this as FIXED once bug 854263 is merged to m-c.)