28910 - Thread view looks different than 4.x - replies not sorting correctly

Status: VERIFIED WORKSFORME

(SeaMonkey :: MailNews: Message Display, defect, P3)

Description

[fenella]

Linux (2000-02-22-12 M14)
Win32 (2000-02-21-09 M14) (win32 2/22 build is broken)
Steps:
1. Launch Messenger using -mail option
2. Select a message that has 5 replied message
3. View this message in thread view.
Actual result: (message are numbered in the order of replies)
>msg 1
   >reply 5
   >reply 6
   >reply 7
>reply 2
   >reply 3
Expected result: Like nova, it should be like this:
>msg 1
   >reply 2
   >reply 3
   >reply 4
   >reply 5
   >reply 6
 This occurs on Linux and win_nt 4.0

Comment 1

[Richard Zach]

I think the problem is that we now use threading by reply-to headers instead of
by subject.  This works in News, but in Mail, this will often be broken, because
the Mail thread often doesn't have your replies in them (they're saved in the
Sent folder). Bug 18789 is an RFE for making this a pref.  But the standard
behavior should should be threading by subject first, then by reply-to headers. 
I think bienvenu is the threading person...

Comment 2

[lchiang]

Fenella, what are the subjects of your msgs?  Are they all the same?

Comment 3

[Richard Zach]

I'm seeing this too.  They all have the same Subject, but are threaded by
Reply-to and X-Reference headers. Attaching a testcase.

Comment 4

[Richard Zach]

Attachment: Mailbox file being threaded incorrectly

Comment 5

[David :Bienvenu]

we should also be threading by subject. If that's broken, then threading is 
broken for most people. Sigh.

Comment 6

[David :Bienvenu]

damn, if the reply starts with Re:, it doesn't get threaded by subject, which is 
wrong. Maybe the Re: is not getting stripped off when we're matching on subject. 
Argh.

Comment 7

[fenella]

To answer Lisa's question. Yes I have the same subject.

Comment 8

[leger]

Need more info on why this broken, risk of fix, etc.

Comment 9

[David :Bienvenu]

Why this is broken? At first glance, because I suck. I can't see what the code's 
doing wrong and indeed, this seems to work sometimes. If this bug is not fixed, 
my life is going to be a living hell during beta because people will complain 
about it ALL THE TIME, if 4.x betas are any indication.

Comment 10

[Richard Zach]

David: I don't know the code, obviously, but it looks like it works if there is
no In-Reply-To header.  Are you threading on subject only if there are no
In-Reply-To or References headers?  Can you just turn off threading by headers
in Mail folders?

Comment 11

[David :Bienvenu]

we currently ignore In-Reply-To, so I doubt that's it :-) I always thread on
subject unless you've set a pref to turn it off. I saw this bug happen last
night, and I don't have the pref set, so I don't think the pref is involved. And
the pref only turns off threading message with a common subject when the subject
in question does NOT start with Re: - in your example the mis-threaded message
starts with Re:

Comment 12

[Richard Zach]

Remember bug 20421?  That was a similar, for empty subject.

Comment 13

[David :Bienvenu]

Richard, do you have a local mail folder that exhibits this problem when you
delete the msf file? That would make it easy for me to recreate this.

Comment 14

[Richard Zach]

I'm way ahead of you ;)
Testcase is already attached.

Comment 15

[Richard Zach]

Attachment: Screenshot of incorrectly threaded mailbox, for good measure

Comment 16

[lchiang]

I'll clear the Need Info status so PDT team can evaluate and put PDT+ on this.

Comment 17

[David :Bienvenu]

In the attached test case, the subjects are not identical. The first subject is 
unique, the second subject occurs three times, etc. I'm not sure if it's a bug 
in copying the text of the attachment from a browser window or not, but Richard, 
if you could stare at them a bit, I'd appreciate it. For example, in the first 
subject, there are three spaces before the J in Jamming, but two spaces before 
the J in the second subject.

Comment 18

[David :Bienvenu]

when I made all the subjects identical, this threaded fine. So I don't know how 
to recreate this. I swear I did see some incorrect threading a couple nights ago 
but when I tried to recreate it last night, I couldn't.

Comment 19

[Richard Zach]

I sincerely doubt any of my correspondents inserted the spaces by hand.  So I
suspect it's a problem with long subject lines, where MTAs or MUAs insert
whitespace (e.g., wrap at 72 chars and then the reply has the CR/LF removed
again).  Will investigate. fenella, do the messages you have problems with also
have long headers?

Comment 20

[David :Bienvenu]

that makes a lot of sense, since the differences always start at column 72. My
guess is that 4.x threads this folder the same as 5.0.

Comment 21

[David :Bienvenu]

whoops, except that 4.x only considered the first 50 or 60 characters or so when
matching subjects, IIRC. But that's a bug, sorta.

Comment 22

[Richard Zach]

Attachment: Another testcase with non-ascii characters in the header

Comment 23

[Richard Zach]

I attached another mailbox where I sent a message with non-ascii characters in
the header, and then replied using different mailers and different settings
(quoted-printabl on/off, etc.).

Comment 24

[lchiang]

fenella is out today unfortunately.  She can attach her mailbox file on Monday.

Comment 25

[leger]

PDT will wait until Monday to make call on this.  Need fenellas info.

Comment 26

[fenella]

Attachment: Mailbox contains replied messages

Comment 27

[David :Bienvenu]

that attachment just seems to contain one message, so it's  not useful to me.

Comment 28

[fenella]

Attachment: Please use one. It is a local folder with test messages

Comment 29

[fenella]

Sorry. I lost the original messages.
So I re-created some replied message using the same subject and different 
subjects.
Read them using Linux (2000-02-28-08 M15) and win32 ( 2000-02-28-09 M15) builds, 
I do not see the same problem any  more.

Comment 30

[David :Bienvenu]

so, should we mark this works for me? (not worrying about the various wrapping 
that various MTA/MUA's are doing to the subjects)

Comment 31

[fenella]

Zach, to answer your question, no, mine message does not have long headers.

David, in my opinion, since the original problem cannot be recreated, we can
mark this worksforme.  And if Sach has concern on longer headers (or MTA or MUA)
issues, he can open a separate bug.

Comment 32

[leger]

Per Phil, marking WorksForMe.

Comment 33

[leger]

Marking Verified.