<html>
<head>
<meta http-equiv="content-type" content="text/html; charset=utf-8">
</head>
<body bgcolor="#FFFFFF" text="#000000">
Sending this to firefox-dev for lack of a more-specific mailing list
specifically about e10s.<br>
<br>
Below is a list of the nightly topcrash signature specifically
related to e10s being enabled, along with some notes about related
bugs:<br>
<br>
<b>#1: mozalloc_abort(char const* const) | NS_DebugBreak |
mozilla::ipc::MessageChannel::DebugAbort(char const*, int, char
const*, char const*, bool) |
mozilla::ipc::MessageChannel::Send(IPC::Message*, IPC::Message*)</b><br>
A bunch of bugs are filed for this crash signature, including
1047563, 1041457, 1042786, 1021053, 1046478, 1056018, and 1046426<br>
<br>
It seems that this signature is too generic, and we should break it
down in more detail based on the abort message and perhaps other
useful data. blassey/jim, is there somebody from the e10s team who
can help figure out a better signature algorithm? Once there is an
algorithm, getting it implemented on crash-stats should be pretty
trivial.<br>
<br>
<b>#2: CrashReporter::GetExtraFileForMinidump(nsIFile*, nsIFile**)</b><b><br>
</b>Bug 1050210 - plugin refactoring *should* fix this, but it seems
like we could also band-aid patch it for now to stop the crashiness.<br>
<br>
<b>#3: mozilla::OffTheBooksMutex::Lock()</b><br>
Bug 1067348 - related to the above bug, apparently<br>
<br>
<b>#4: mozilla::layers::ShadowLayerForwarder::IsSameProcess()</b><br>
Bug 1068626 - has STR<br>
<br>
<b>#5: imgRequest::GetURI(mozilla::image::ImageURL**)</b><br>
Bug 1068652<br>
<br>
<b>#6: libc-2.11.1.so@0x335a5<br>
</b>The stack is so badly broken here that I'm not sure we can do
much with this one. It's likely to be a Linux-specific variant of
something above.<br>
<br>
<b>#7:
js::jit::LInstructionVisitor::visitModPowTwoI(js::jit::LModPowTwoI*)</b><br>
I really don't understand why this is e10s-specific, but it does
appear to be! No bug filed yet, but at Crashkill we'll get one
filed.<br>
<br>
<b>#8: mozilla::net::HttpChannelParent::FailDiversion(tag_nsresult,
bool)<br>
</b>Also no bug filed yet.<br>
<br>
<b>#9: mozilla::jsipc::ObjectId::ObjectId(unsigned __int64, bool)</b><br>
Bug 1075543, with firebug, STR in bug 1077587<br>
<br>
<b>#10:
mozilla::layers::CrossProcessCompositorParent::ForceComposite(mozilla::layers::LayerTransactionParent*)<br>
</b>No bug filed yet.<br>
<br>
There's a long tail, but these are the ones in priority order that
are most important. <br>
<br>
Here is the supersearch query I'm using to produce this
e10s-specific topcrash list (view the signature facet):<br>
<a class="moz-txt-link-freetext" href="https://crash-stats.mozilla.com/search/?dom_ipc_enabled=__true__&release_channel=nightly&build_id=%3E20141001000000&_facets=signature&_columns=date&_columns=signature&_columns=product&_columns=version&_columns=build_id&_columns=platform">https://crash-stats.mozilla.com/search/?dom_ipc_enabled=__true__&release_channel=nightly&build_id=%3E20141001000000&_facets=signature&_columns=date&_columns=signature&_columns=product&_columns=version&_columns=build_id&_columns=platform</a><br>
<br>
And then comparing against the non-e10s topcrash list (which I can't
link to because of a limitation in supersearch).<br>
<br>
--BDS<br>
<br>
</body>
</html>