256332 - RSS breaks after compact folder

Status: RESOLVED FIXED

(MailNews Core :: Feed Reader, defect)

Description

[Adrian Moisey]

User-Agent:       Mozilla/5.0 (Windows; U; Win98; en-US; rv:1.7.2) Gecko/20040816 Firefox/0.9.1+
Build Identifier: Mozilla/5.0 (Windows; U; Win98; en-US; rv:1.7.2) version 0.7+ (20040820)

RSS breaks after you compact folder on one of the feeds

Reproducible: Always
Steps to Reproduce:
1. Subscribe to a feed
2. Right-click on it, click Compoct Folder
3. Attempt to get feed working (click "Get New Messages")

Actual Results:  
RSS doesnt work

Expected Results:  
RSS should update when new items are posted

Comment 1

[David :Bienvenu]

ah, probably losing the feed uri after the compaction. I can look into this.

Comment 2

[Scott MacGregor]

could be a good 0.8 blocker

Comment 3

[David :Bienvenu]

I've coded up a fix for this...it's not quite done, but basically it involves
saving all the db folder info from the original db and storing it in the new db
generated by compacting the folder.

Comment 4

[Scott MacGregor]

Thanks for working on this David. Any chance the patch will also have a side
effect of making us rememmber the feed url when rebuilding the summary file for
cases like: Bug #255453?

Comment 5

[David :Bienvenu]

yes, if we do rebuild the summary file in that scenario, instead of just
noticing that the summary file is out of date and erroring out...

Comment 6

[David :Bienvenu]

Attachment: proposed fix

this changes the way saving and restoring of dbfolderinfo's works - it saves
all the data blindly, which is nice because when new data is added, we don't
need to add new code to save+restore it, and it cuts down on code size. The
downside is that there's some data we don't want to restore, like timestamps,
so that data we have to explicitly set after restoring the db folder info.

I've also removed some unrelated trace statements that NS_ENSURE_TRUE was
generating.

Comment 7

[Scott MacGregor]

(In reply to comment #5)
> yes, if we do rebuild the summary file in that scenario, instead of just
> noticing that the summary file is out of date and erroring out...

Actually I think what we saw was that the summary file was out of date and it
errored out. So this probably won't fix that particular case. 

Thanks for working on this.

Comment 8

[Scott MacGregor]

Comment on attachment 156812 [details] [diff] [review]
proposed fix

1) These lines:

if (err == 0)

can just be

if (!err)

2) I assume that mProperties contains non string properties such as sort order,
sort type, view flags, etc. so we aren't losing those non strings during
compaction? 

3) I didn't see why LoadMemberVariables() now gets called in
InitFromTransferInfo, does that re-load the view flags, sort oder, etc. from
the property values we just read back in?

Comment 9

[David :Bienvenu]

1) I'll fix that.
2) everything in mork is a string - it's a text format db.
3) Yes, LoadMemberVariables gets the data back from the mork row into the member
vars of the dbfolderinfo.

Comment 10

[Scott MacGregor]

We can close out this bug now right David? Looks like it's checked into the
branch and trunk.

Comment 11

[David :Bienvenu]

yep, sorry. Marking fixed.

Comment 12

[Adrian Moisey]

You guys rock! Open Source rocks!