1272151 - Crash in IPCError-browser | (msgtype=0x220003,name=PBlob::Msg_PBlobStreamConstructor) Processing error: message was deserialized

Status: RESOLVED WORKSFORME

(Core :: Networking: HTTP, defect, P1)

Description

[Jim Mathies [:jimm]]

This is showing up in beta 47 under e10s in 

mozilla::net::HttpChannelChild::AsyncOpen(nsIStreamListener*, nsISupports*)

Comment 1

[u408661]

Jason - so far, we're only seeing this on 47 (it seems). Do we have someone who can look into this? (Not sure how urgent it is, given it doesn't appear to be showing up on aurora or beta, and e10s is, iirc, not continuing to release with 47.)

Comment 2

[Jason Duell]

> > This is showing up in beta 47 under e10s
>
> so far, we're only seeing this on 47 (it seems)... it doesn't appear to be showing 
> up on aurora or beta

Did you mean to say "aurora or nightly"?  Because beta == 47, right?

If it's showing up only for e10s in 47 and not in <=46 or 48/49, then yeah, it's not going to be top priority.

Comment 3

[Jim Mathies [:jimm]]

The most likely reason we don't see it on nightly or aurora is user volume. We have about 2 million users on beta using e10s right now where as nightly and aurora are probably less that 50K.

My interest in getting this looked at is that it's an e10s only crash (currently #39). e10s crash numbers are the only thing blocking rollout currently.

https://crash-stats.mozilla.com/search/?product=Firefox&version=47.0b1&version=47.0b2&version=47.0b3&version=47.0b4&dom_ipc_enabled=!__null__&_facets=signature&_columns=date&_columns=signature&_columns=product&_columns=version&_columns=build_id&_columns=platform#facet-signature

Comment 4

[Jim Mathies [:jimm]]

Sometimes these errors are easy to fix, the complete ipc error string is:

(msgtype=0x220003,name=PBlob::Msg_PBlobStreamConstructor) Processing error: message was deserialized, but the handler returned false (indicating failure)

So basically the constructor on the parent side returned false indicating a failure - 

http://mxr.mozilla.org/mozilla-central/source/dom/ipc/Blob.cpp#4087

One thing you can do is modify these returns such that they propagate the failure back to the child where it gets handled.

Comment 5

[u408661]

(In reply to Jason Duell [:jduell] (needinfo? me) from comment #2)
> > > This is showing up in beta 47 under e10s
> >
> > so far, we're only seeing this on 47 (it seems)... it doesn't appear to be showing 
> > up on aurora or beta
> 
> Did you mean to say "aurora or nightly"?  Because beta == 47, right?

Yes. Too many words, not enough brain :D

Comment 6

[Jason Duell]

Nick: can you see if this is still 47-only, and what we should do here?

Comment 7

[u408661]

According to https://crash-stats.mozilla.com/signature/?date=%3E%3D2016-06-02T17%3A17%3A00.284976&date=%3C2016-06-09T17%3A17%3A00.284976&signature=IPCError-browser+%7C+%28msgtype%3D0x220003%2Cname%3DPBlob%3A%3AMsg_PBlobStreamConstructor%29+Processing+error%3A+message+was+deserialized this is still 47-only. However, given 47 was only promoted to release this week (and therefore 48 only promoted to beta this week) let's give it another week to see what happens on beta with its larger population than aurora. If it's still 47-only next week, I think we can just close this.

Leaving the ni? so I remember to come back to this in London.

Comment 8

[Jim Mathies [:jimm]]

Note the hexadecimal message ids in these signatures can change across builds / releases. Use a token search to find all occurrences - 

https://crash-stats.mozilla.com/search/?product=Firefox&signature=~Msg_PBlobStreamConstructor&_facets=signature&_columns=date&_columns=signature&_columns=product&_columns=version&_columns=build_id&_columns=platform#facet-signature

Comment 9

[u408661]

So using the search from comment 8, it still looks like these are only 47 beta and earlier. I'm going to mark this WFM. Let's re-open (or open a new bug) if the Msg_PBlobStreamConstructor signature appears again.