Hi David,
* David Bremner <david@tethera.net> [26. Jan. 2014]:
> Gregor Zattler <telegraph@gmx.net> writes:
> 
>> Today I produced another mbox with the very same command but with
>> a now larger email corpus freshly indexed with a fresh notmuch.
>> The mbox contains (according to mutt) 507 messages in 34 threads.
>> One of them is the thread I searched for.
>>
>> I grepped for the 7 subjects within the 34 subjects and only 5
>> showed up.
> 
> I don't know what you mean here. Grepped where? in the raw messages?

With mutt I had a view at the collapsed 7 respective 34 threads.
One then sees the very first E-Mails of a thread and among other
information their subjects.
Via editing I produced two lists with subjects and then searched
each of the 7 in the list with the 34.

>> If somebody want's to dig into this: I can provide the two
>> mboxes. 
>>
>> Disclaimer: Many of the emails which arrived before the problem
>> report are not the exact same than then, because since the I
>> mangled them with a script.  This should have not changed the
>> threading but I cannot be 100% sure.  But if it's important for
>> further investigation I'm probably able to reproduce the status
>> quo of the email corpus then from my backups.
> 
> If it's currently not working then I guess your current corpus should be
> fine. It would probably help to restate what exactly is wrong. There was
> a lot of discussion, and the concrete problem I saw identified (in
> id:874nvcekjk.fsf@qmul.ac.uk ) was that certain malformed In-reply-to
> headers were causing unrelated threads to merge.

Yes.  I understood the commit message of the commit you
referenced in the email I answered to, that now notmuch uses
 Reference: headers to do the threading.  I had a quick view at
the References header in the mbox file and none looked
suspicious.

Ciao, Gregor
-- 
 -... --- .-. . -.. ..--.. ...-.-