I did manage to solve the previously mentioned interpolation bug, but ended up completely sidetracked into refactoring the entire system of interpolated rendering, hence the late update now.
I think I just have one tiny issue to fix, regarding auto-centering, and then interpolation is done: Basically, there is the occasional situation where the screen is redrawn in between a device’s location moving and the screen being auto-centred, producing a slight jump effect.
I will get this sorted later today by synchronising all selected device location changes with auto-centering.
Aside from that, I just need to diagnose and fix an issue where zoom level change detection is broken and then I can draw a line under this phase.
I will return to work on GeeOhTileProcessor – a tiny refactor for handling error responses better – and then do a “final” pre-loading run, before coming back to the mobile app and finishing it off.
Oh wow, I’ve been wrestling with such an annoying bug in trying to implement full interpolated movement for multiple devices in GeeOh! Mobile, and every time I think I’ve cornered it, it seems to morph into something else.
Been chasing my tail for days now.
The interpolation works fine for selected device, but any other device’s interpolated rendering goes badly wrong, quickly.
I feel like I’m really close now, maybe another day or so, but I could be wrong. I shall be back on here writing a celebratory post once I’ve caught the little bastard.
So, as previously hinted at, I have been working on implementing “always on” interpolation (toggleable in the options menu) – which I have already managed for “this” device – and I have nearly got it sorted for all external devices too.
I’ve just been trying to make the Interface as efficient as possible, not overdoing the number of methods etc. required, but I think I’m just going to bite the bullet and make getters/setters for all the essential variables.
This will be the last “big job” I do before Gee-Oh! Mobile is ready for release. I think I will have a few business decisions to make before then (ie. dividing the business up into smaller companies – one per product – to remove confusion for potential investors etc.), but as far as technical decisions and implementations go, I’m just about done.
I also did some improvements to the map tile pre-loading application (and also to the memory management, how it’s deployed etc.) and I’m now estimating just a further 10 days before the main map tile server pre-load is complete.
Edging ever closer to getting this shit done!
The cheesy title suggests I’m starting to blossom into a caring, generous pillar of the community. Nope 😉 The actual purpose of this post is to talk about the work I’ve been doing in allowing auto-centre to be applied to external devices – not just the device the app is running on – in Gee-Oh! Mobile.
This is a pre-cursor to full integration with Gee-Oh! Server, allowing full tracking and communication between networks/groups of devices. Exciting shit.
Now let’s get into the boring technical detail…
I’ve had to literally rip the guts out of the main class, with all its references to global variables etc. and update this with getters/setters, kept in check by adding additional method references to the device interface. While I “had the bonnet up”, I did a ton of additional fixes, and it ended up being probably the biggest single commit I’d ever be comfortable making, but there you go.
Very pleased with the way it has gone, though. I even snuck in a little “loading” graphic to remove the jerkiness while we wait for interpolation to kick in.
It’s these little details that will make all the difference.
Let’s have a little recap on all the remaining tasks before release:
- Finish pre-loading server with UK data
- Confirm that all “centre on selected device” work is done for now
- Including making sure “test flight” debug mode works with replay etc.
- In actual debug mode, substitute replay or flight devices with “this” device
- Highlight selected device/waypoint, whether or not we have GPS location for “this” device
- Refactor selected waypoint logic to be the same as for selected device
- Look at implementing “always on” interpolation
- Move point pair stack into individual devices
- Each device must have its own interpolation thread running
- Must retain global control of this
- Shrink waypoints as we zoom out
- Replace current waypoints menu with proper pop-up menu
- Look at implementing “proper” double buffered rendering
- Make placenames updatable via server API
- ProGuard builds
I think that just about covers it. I’m not going to add any more features, I just need to clear this list and Gee-Oh! Mobile is getting shoved out the door.
Can’t happen soon enough.
Just a vague update on the commencement of Gee-Oh! Mobile development. Since I got the server back on track and processing squares again last week, I’ve dived right back into the world of Android.
I managed to score a few quick wins from my previous “barriers to release” list, including the rendering of placenames (fading in and out accordingly), and also added the fading out of motorway/main road labels as we zoom out to particular values.
I did a little bit of housekeeping, removing a superfluous ViewThread class that was never really being used as intended, and lots of other little bits, too many to remember fully. Been a busy boy.
My latest venture is a very large refactor to incorporate multiple devices (for now, only a replayDevice and the debug testFlightDevice) and allow auto-centering on these also whilst preserving our actual location indicator on screen. I’ll do a separate blog on this, actually – ’tis a big ol’ topic.
And so this ends.
This was actually complete over a week ago, I’ve just not got round to writing about it until now. Soz.
I switched my hosting from a VPS to a Cloud Server, still with 1&1, to get the benefit of (practically) unlimited inodes – as described in my last blog entry – which required setting up everything from scratch (including installing Java!) on the new platform and then copying across anything relevant from the VPS.
The biggest difficulty I came across was in sending notification emails. The main issue from my end was that another incorrectly setup service was already running on port 25 (so I found that and killed it), but the real show-stopper was with 1&1 – who were blocking all outgoing traffic to port 25. I had to get support to remedy that for me.
Aside from that, all seems good. I will have to get onto 1&1 soon and ask them to kill off the VPS (and its associated bill, of course).
I was able to resume pre-loading of GeeOhTileServer grid squares pretty quickly, and that now continues apace. Still on course for mid-June – ’tis a slow process.
Had a bit of a nightmare over the last week or so. I’d been running the map tile pre-loading program for a while when all of a sudden it just stopped working. I had to rewrite parts of one of my Java libraries in order to catch the error in more detail, and it was coming back as “disk full”.
When I checked on my server’s control panel, it was showing that I had reached my 500,000 inode limit. Yeah, I didn’t know what an inode was either. Turns out it just means file, and that my VPS was limited to half a million files – that’s no good when I need at least 597871 map images (or “empty” files to indicate “all water” or “all land”), plus two meta files for each.
It took me a while to get some sense out of my hosting provider, but eventually I was informed that I need to migrate to a Cloud Server (which, I was assured, have absolutely shitloads of inodes).
I double-checked that I could upgrade to a better Cloud Server package when I needed to scale, without having to re-install everything, and was given some reassurance that they just have to provision more resources. Sound.
So, this week I will be provisioning a shiny new Cloud Server and attempting to just scp everything across from my existing VPS.
Should. Go. Smoothly.