On 02/10/2017 00:38, Ben Bucksch wrote:
You already did most of the investigation work with:
replace the original message file fails with "access denied" since due to the new processing the file is still locked.
You might just need to close the streams (earlier), or something like that.
I also see some disconcerting messages come from Necko, like: WARNING: 'NS_FAILED(rv)', file c:/mozilla-source/comm-central/mozilla/xpcom/io/SlicedInputStream.cpp, line 119
The warnings might give additional hints what's going wrong.
I did indeed do some initial investigation to conclude that this is a hard problem which might involve fixing difficult timing conditions in Necko code. (More than eight hours work on a Sunday are not possible.)
Given my current workload, I can only put this on the back-burner. There is a simple patch to fix the problem immediately, and anyone who wants to investigate further can simply backout that patch locally and do their investigation without any stress, while the user community can enjoy a working Daily.
Let me quote a comment I read on BMO: "It may well be that sliced[*] as implemented for Fx is no good for Tb purposes, and those driveby no-fork critics either figure it out themselves or pipe down, as a fork may unfortunately be necessary."
Jörg.
[*] SlicedInputStream
Jörg Knobloch wrote on 02.10.17 01:48:
(More than eight hours work on a Sunday are not possible.)
Nobody expects you to fix every possible problem within a few hours on
the same day.
There is a simple patch to fix the problem immediately
It's not a "simple patch". Forking a core Necko file is not trivial,
simple or cheap.
On 02/10/2017 02:12, Ben Bucksch wrote:
Nobody expects you to fix every possible problem within a few hours on
the same day.
But you do expect me to fix every possible problem, right? If not on the
same day, then later, right?
Jörg.