114501 - Mail folders are always refreshed on OS/2

Status: VERIFIED FIXED

(SeaMonkey :: MailNews: Message Display, defect)

Description

[Julien Pierre]

If you have a large Inbox mail folder, this is a very annoying bug :
1. start mozilla messenger
2. mozilla builds the Inbox folder summary file when first accessed
3. exit mozilla
4. restart mozilla messenger5. mozilla again rebuilds the folder summary file ,
even though there is no need as no database corruption happened

This applies to all mail folders, not just Inbox. It causes the mail application
to be very slow on OS/2.
Mike Kaply of IBM helped me debug this over IRC today and it turned out there is
a date problem with time_t being used. On OS/2 time_t is a double and that
caused problem in the code - I guess it thought the summary file was always out
of date with the folder.

Comment 1

[Mike Kaply [:mkaply]]

Attachment: Change time_t to PRUint32

On OS/2, time_t is a double (you don't want to know - see bug
http://bugzilla.mozilla.org/show_bug.cgi?id=104047)

anyway, the msgdb code uses time_t for the mFolderDate here -
http://lxr.mozilla.org/seamonkey/source/mailnews/db/msgdb/public/nsDBFolderInfo.h#109


But honestly, I don't think it should.

If you look through the various code that access mFolderDate, it treats as a
PRUint32 explicitly (including some casts)

So the fix is this:

< time_t      m_folderDate;
> PRUint32	m_folderDate;

I've rebuilt mailnews with this and had no problems. It doesn't appear that
mailnews uses mFolderDate for any actual C runtime time/date calls - it only
uses it for storage.

Comment 2

[David :Bienvenu]

Comment on attachment 61180 [details] [diff] [review]
Change time_t to PRUint32

r=bienvenu

Comment 3

[(not reading, please use [e-mail] instead)]

Comment on attachment 61180 [details] [diff] [review]
Change time_t to PRUint32

sr=sspitzer

I guess it wasn't nsFileSpec, my gut lied to me!

Comment 4

[Brendan Eich [:brendan]]

a=[e-mail] for checkin to the 0.9.7 branch.

/be

Comment 5

[Mike Kaply [:mkaply]]

Fix checked into trunk and 0.9.7 branch

Comment 6

[esther]

julien, can you verify this is fixed since it appears to be reproducible on IBM
OS/2.  I did not see the problem on my WinXP and I don't have access to an OS/2
system.  

Comment 7

[Julien Pierre]

Yes, I verified it on a couple of my builds in the last few days. Marking verified.