397910 - Message Drafts Not Saved When Composed Offline

Status: RESOLVED FIXED

(MailNews Core :: Networking: IMAP, defect)

Description

[Scott MacGregor]

Using a trunk nightly.

1) Go Offline
2) Compose a new message but don't send it & don't save it, leave the compose window open.
3) go back online.
3) Several hours after going back online, something caused Thunderbird to hang

On restart, I was unable to find a draft (Imap or in a local folder) for my message.

Comment 1

[Wayne Mery (:wsmwk)]

upon going online, I received and Alert "The current command did not succeed.  The mail server responded:Empty message to APPEND". Nothing in error console.

Mozilla/5.0 (Windows; U; Windows NT 5.1; en-US; rv:1.9.1pre) Gecko/20090619 Shredder/3.0b3pre

Comment 2

[Jens Müller (:tessarakt)]

Looks similar to bug 433806.

Comment 3

[WADA:World Anti-bad-Duping Agency]

xref some bugs for ease of tracking.

Comment 4

[WADA:World Anti-bad-Duping Agency]

Phenomenon I observed with 20091112 build using Gmail IMAP was Bug 528669 comment #1.

Comment 6

[WADA:World Anti-bad-Duping Agency]

Phenomena observed with Tb 3.0xpre or later in some bugs, with IMAP log or
with detailed observation, instead of steps to reproduce and incident of draft
is lost only. (they ate put in dependency tree)
> (i)   with Gecko/20090619 Shredder/3.0b3pre : Bug 397910 comment #1
> (ii)  with Gecko/20091112 Shredder/3.0pre   : Bug 528669 comment #1
> (iii) unknown, reported on 2010-01-25       : Bug 530709 comment #19

Comment 7

[WADA:World Anti-bad-Duping Agency]

IMAP log with next parameter is required.
> https://wiki.mozilla.org/MailNews:Logging
> NSPR_LOG_MODULES=imap:5,imapoffline:5,timestamp

Comment 11

[WADA:World Anti-bad-Duping Agency]

bug 566655 comment #5 (2010-05-18) and bug 566655 comment #7 (2010-05-19) looks behaviour of current nightlies. (i.e. it looks problem is already resolved)

Comment 12

[WADA:World Anti-bad-Duping Agency]

Value in "Order Received" column is UID of mail when IMAP folder. Show "Order Received" column when you test, please. See Bug 528669 comment #1 for "Order Receved" column value example.

Comment 13

[David :Bienvenu]

I was unable to reproduce this on my Mac a couple days ago, but I'm seeing it in my Windows profile right now, on the same server that worked on the Mac. So I suspect this is more likely to be profile-specific than server-specific (or specific to certain steps that we don't have nailed down yet).

Comment 14

[David :Bienvenu]

What I'm seeing on Windows is my favorite nsILocalFile caching of stat results, which means we think the temp file is empty. So I need to clone the file in order to fix this.

Comment 15

[David :Bienvenu]

Attachment: fix

this fixes the issue. I'm having quite a bit more trouble getting a unit test for this.

Comment 16

[David :Bienvenu]

Attachment: fix with unit test

this adds a unit test for copyFileMessage offline (which is what save draft uses). I had to change a couple assertions to warnings to get the test to succeed w/ the fix.

Comment 18

[Mark Banner (:standard8)]

1.9.0.* have nothing to do with any released TB or SM version.

Comment 19

[David :Bienvenu]

fixed on trunk - http://hg.mozilla.org/comm-central/rev/efe88f656eee

Comment 20

[Mark Banner (:standard8)]

Landed on 1.9.2 branch:
http://hg.mozilla.org/releases/comm-1.9.2/rev/f47d3e48b1a6
http://hg.mozilla.org/releases/comm-1.9.2/rev/a44846d81ebb

Comment 21

[Robert Kaiser]

We won't block or hold off a security update for something that has never worked in this release series. Surely would be nice to take if mailnews people decide so, but not blocking releases.

Comment 22

[Mark Banner (:standard8)]

(In reply to comment #21)
> We won't block or hold off a security update for something that has never
> worked in this release series. Surely would be nice to take if mailnews people
> decide so, but not blocking releases.

Although this patch applies to comm-1.9.1 there's a bit of fuzz (I didn't look in detail at that). The unit test does however just hang on my Mac.

Thunderbird is already upgrading most users to 3.1, so we're not going to block a 3.0.x release on it.

If someone wants to fix this up so that it works on the 1.9.1 branch, then we would consider accepting a patch.

Comment 23

[Thomas D. (:thomas8)]

I'm confused about the status of this bug, because the target milestone is 3.3a1 but it is also marked fixed, and Dan somewhat abiguously said in http://getsatisfaction.com/mozilla_messaging/topics/why_do_i_keep_losing_draft_e_mailslost_e_mail_drafts#reply_2828247 that "one bug... was fixed *in* Thunderbird 3.1". Is this bug really fixed in (i. e. "for") the current release version, 3.1.3?

We need to communicate much more clearly about this bug, users of 3.0.x are getting really angry about this in GS, plus the added confusion of

Bug 497952 - Activity Manager / statusbar message "Deleted x message(s) from [Drafts...]" can cause user confusion (users think there is dataloss)

http://getsatisfaction.com/mozilla_messaging/tags/bug_397910
http://getsatisfaction.com/mozilla_messaging/tags/bug_497952

Users are mixing up the two issues, really big confusion on GS, and many reports for both of them. Users who really have dataloss because of this bug have commented in topics about the bad status bar message (which is harmless), and users who saw the harmless status bar thing think they have lost messages (but they haven't). I've tried to tag as many topics as I could find about this, sometimes with both bugs so that we have easier reference. Someone needs to do some merging in getsatisfaction.

Comment 24

[Mark Banner (:standard8)]

(In reply to comment #23)
> I'm confused about the status of this bug, because the target milestone is
> 3.3a1 but it is also marked fixed ... Is this bug really fixed in
> (i. e. "for") the current release version, 3.1.3?

Target milestone shows which release it is aimed for - or in this case, fixed in. However, that's just the next trunk release.

Look at the status-thunderbird3.1 flag - that says rc1-fixed. So this bug was fixed for 3.1rc1 i.e. 3.1 final.

Comment 25

[Thomas D. (:thomas8)]

We need to flag one of the GS topics canonical for this bug: I recommend http://gsfn.us/t/po2n for that.

http://gsfn.us/t/po2n has a much better / clearer description with detailed STR, explicit version no etc. Therefore, it should be made canonical, and get a decent company comment pointing to this bug and that it's fixed for 3.1.

After that, http://gsfn.us/t/keyv should be merged with the canical one.

Any disagreement about that procedure?

Comment 26

[Dan Mosedale (:dmosedale, :dmose)]

A fine choice; looks like Wayne has helped make it happen already.  Thanks, Thomas!

Comment 27

[Wayne Mery (:wsmwk)]

finishing this up in getsatisfaction. http://gsfn.us/t/po2n is canonical

Comment 28

[Dan Mosedale (:dmosedale, :dmose)]

W00t; thanks Wayne & David! It's great seeing stuff like this solved end-to-end.