Blog
Page: 0 ... 5 ... 10 ... 15 ... 20 ... 25 28 29 30 31 32 33 34 35 36 ... 40 ... 45 ... 50 ... 55 ... 60 ... 65 ... 70 ... 75
Scribe v2 Progress
Date: 17/3/2008
I've ported the existing v2 to both Linux and Mac, although I have no immediate plans to make releases on those platforms I like having a working svn. The linux build is particularly useful for checking under valgrind and the mac port is useful for when I'm working on a laptop away from my PC.

As for v2 itself I've been implementing the foundations for a basic Maildir backend which I can reuse for IMAP. I know its not the "proper" way you should do IMAP but I can move towards that once I've got something working. I'm putting in place methods of converting mail between different backends and making that fairly seemless. Today I've implemented an attachment class for the IMAP code and I'm using the GMime class that I wrote a zillion years ago.

Because the new internal API's are now well designed I can change implementation details like exactly how the IMAP cache works without effecting the main app in any significant way. This gives me a lot of freedom that I didn't have before.

Some of the work I've been doing this week is going towards adding multiple simultaneous folder support. Which is a stepping stone to good import and export functionality too. I could do something crazy like write a MAPI backend... thing about that. Scribe could be corrupted by the dark side!

If I wrote a backend for all the mail import/export formats I can read/write, and then rewrite the import/export code to replicate between any 2 backend mail stores... hmmmmm. The possibilities are endless.

So far the whole architecture is holding up well in day to day use. I can't wait to write a faster "native" backend to replace mail2 files.
(1) Comment | Add Comment

v2 Where For Art Thou?
Date: 5/3/2008
So I guess people are wondering where v2 has gone? Well works continues, abet at a slower pace due to life's interruptions. Currently I'm ripping out the old cruddy IMAP full implementation and making a new one based on the new API's that I've made. This is currently looking a lot like a MAILDIR implementation where all folders and email are read from disk. Basically my idea is to cache email and folders on the IMAP server on disk in a sub-folder under the Scribe location such that you always have an offline database you can access. The IMAP network thread then attempts to keep the folders in sync with whats on the server. I haven't got to that yet, because I'm still building the MAILDIR functionality up. However I get a free MAILDIR implementation at the end as well.

Anyway I might make the implementation of the caching/offline part different later as I build more optimal backends but MAILDIR is way easy and fast to implement (at the expense of runtime performance) so it'll get me a working IMAP implementation faster. And I think everyone would prefer working IMAP sooner even if it's a little slower than it could be. I'm not planning to complicate the MAILDIR stuff with indexing but that would probably fix the speed issues at the cost of extra complexity and implementation time.

This will hopefully make the Scribe v2 upgrade worth it.
(2) Comments | Add Comment

The evils of cidaemon.exe
Date: 22/2/2008
I have been regularly having issues writing to files. The symptoms would basically be, try and open a file, OpenFileW would hang for about 2 minutes and then return a valid handle to the file. This causes havoc in my applications as you can imagine. So I fired up the wonderful SysInternals process explorer and looked for what had that file open and some dinky little process called 'cidaemon.exe' has it open.

Dear Google, what is 'cidaemon.exe'?

Dear Matt, 'cidaemon.exe' is the Microsoft indexing service.

Riiiiight. Ok well it's indexing my drive for "searching" which btw sucks on windows. So I'm going to turn it off. You can do that on XP by clicking Start->Search->Files/Folders and then in the Search Options clicking the "Indexing Service" field and it'll give you an option to disable it.

I consider a process regularly holding a non-sharable lock on MY file for 2 minutes severely broken. So I'm going to see what life is like with the indexing service. I suspect that life will be sweeter, my machine faster and my files will open easily.

But this IS windows were talking about... *sigh*
(1) Comment | Add Comment

Scribe v2 Beta?
Date: 17/2/2008
Well due to some timely feedback on the v2.00 alphas we might be approaching beta (as in all the functionality working again) sometime in the next week. I've crossed off all the bugs reported against alpha3 and made a release for alpha4.

Do what do you think? Any good?
(8) Comments | Add Comment

MiserySoft's Finest
Date: 15/2/2008
Re: Visual C++ .NET Can't Find Squat.

Well guess what, the bug is still there in Visual C++ 2005. 2 Major releases and 1 service pack later. And the fix still works as well.

Unbelievable.

So it appears I have to use Visual C++ 2008 now. F. I guess that means I'm looking for a new job. F.

(0) Comments | Add Comment

The Big Scribe Re-write
Date: 8/2/2008
Yes. I'm in rewrite number 4 of Scribe. I've taken an extremely pessimistic view of what I can achieve and aimed low. Let me explain, in all the previous efforts I was trying to replace the backend at the same time as changing the API between the objects in memory and the disk storage interface. This was always problematic because there is too much going on at any one time and it unravels and I lose momentum and the rewrite dies...

So this time, a new idea. Don't re-write the backend, just use the existing mail2 code BUT fix the object's so that they have a clean interface to the storage code. The system I've built on for the last n years has been one with a less than ideal architecture. And now I'm finally ripping that up and replacing it with the new. The in memory objects for Mail, Folders, Calendars, Contacts etc all derive from the "Thing" class (yes I know... original) and they include all sorts of data members that get written to disk. This meant that the "Thing" objects were the only code that knew how to read and write themselves to the mail2 file. The mail2 code didn't know what it was storing. So for me that meant I couldn't read/write the objects to a different back end easily. Now I've separated out the serialization code from the "Thing" objects into a pure virtual API that glues the Things to their mail2 backend. But of course can be replaced with any other backend that implements the glue API.

So that could be a Sqlite database, an MBOX store... or a MAILDIR folder. Or a caching IMAP implementation.

Anyway, currently the app boots, loads a folder and shows email. It's not sending or receiving yet, but I think I'm close to getting that working. Receiving should be easy, sending... uh less so. Because in the object hierarchy shuffle the Mail object lost it's "MailMessage" parent which the sending code uses to encode the outgoing mail. That means more mucking around for me when preparing a message for SMTP. But no doubt I'll have that licked in short order.

One of the releases in the next few weeks will be off this new architecture. And it'll suck. Seriously it'll be a big backward step compared to the latest v1.90 test2 release. Because lots and lots of code has changed... challenging, application wide refactoring. And I won't get that right first go... it'll need lots of field testing, and it won't even do any new tricks to start with either. It'll be the same old UI and feature set... just with more bugs, missing features and crashes. You like? Well it had to be done, and I'm giving the beta testers fair warning that all hell will break loose in the following releases.

So there. More updates to follow.
(10) Comments | Add Comment