Android 2.2 Development

[Guest totiadrenalin]

The Cyanogen vold.fstab is for different hardware. The correct sysfs path for our device is /devices/platform/msm_sdcc.1/mmc_host/mmc0. The vold binary still uses vold.fstab, I'm just hardcoding the partition number (which doesn't work correctly in the aosp implementation).

And what about the jit?

Did you test it?

[Guest totiadrenalin]

And what about the jit?

Did you test it?

flibblesan!

How actually did you fix the marked bug?

I still can't fix it.

[Guest totiadrenalin]

flibblesan!

How actually did you fix the marked bug?

I still can't fix it.

Gallery3D fixed.

Just replace /system/app/Gallery3D.apk with this1Gallery3D.apk

[Guest totiadrenalin]

Gallery3D fixed.

Just replace /system/app/Gallery3D.apk with this1Gallery3D.apk

Onuther Bug!

The videos play's but I can't see anything at all.

[Guest Tom G]

And what about the jit?

Did you test it?

How to test it? The phone hasn't crashed yet.

[Guest totiadrenalin]

How to test it? The phone hasn't crashed yet.

Nice

So it is working well.

Nice to hear that.

[Guest totiadrenalin]

Nice

So it is working well.

Nice to hear that.

And check the videos bug.

Edited September 4, 2010 by totiadrenalin

[Guest totiadrenalin]

And check the videos bug.

Also wifi tethering bug.

Enabling NAT rules...

failed

[Guest Tom G]

I'm just hardcoding the partition number (which doesn't work correctly in the aosp implementation).

It looks like the partition number choice (from vold.fstab) isn't implemented in the aosp code, but it should loop through all available partitions until it finds one suitable. This however is depend on a block uevent value (PARTN) which mustn't be implemented in our kernel (probably a 2.6.32 thing). If the uevent value isn't available it doesn't index the partitions properly (a bug in the aosp code).

I've changed the way it indexes partitions when PARTN isn't available. It should make it work as design whether PARTN is available or not. Its pretty simple really.

Patch for DirectVolume.cpp

135c135

<		 mDiskNumParts = 1;

---

>		 mDiskNumParts = 0;

178c178

<		 part_num = 1;

---

>		 part_num = ++mDiskNumParts;

This way it would still work if part 1 is ext2 and part 2 is fat.

Edited September 4, 2010 by Tom G

[Guest Tom G]

Hurey!

32,23 MFLOP/s

in 0.43 ms

in linpack

Really? I've never played with linpack before, but just ran it with JIT and got 3.796 MFlops/s in 1.43 seconds.

[Guest rulerofkaos]

arm5 isn't supported by JIT (dalvik/vm/Android.mk needs to be modified to remove a check) but its looks like they have had success with it over at xda-dev on other msm7k devices. Its building now (had to do a clean build).

I wonders if compcache is an option on the pulse. I think with such low physical memory it will only make it worse.

PS. I think swap support may be a requirement for compcache so it might not be an option yet anyway.

Tom, can you please upload your latest built with working sensors etc?

[Guest thehobnob]

arm5 isn't supported by JIT (dalvik/vm/Android.mk needs to be modified to remove a check) but its looks like they have had success with it over at xda-dev on other msm7k devices. Its building now (had to do a clean build).

I wonders if compcache is an option on the pulse. I think with such low physical memory it will only make it worse.

PS. I think swap support may be a requirement for compcache so it might not be an option yet anyway.

Just want to say I have Cyanogenmod hacked onto my HTC Kaiser (don't ask, see XDA...:P) with both JIT and 16MB compcache enabled, and that device has a 400Mhz MSM7200 CPU and 128MB RAM, so you should be able to get it to work somehow.

I love what's happening with this so far though!

[Guest FrogPulse]

Are we still using github or are we all doing our own thing?

Looking to doing some coding since my WoW has run out :P

[Guest flip360]

Nice work Tom G, NICE WORK. I would like to try the release with the fixes too, but i will patiently wait for a release :P (hope its quick)

Edited September 4, 2010 by flip360

[Guest Josh04]

Tom, can you please upload your latest built with working sensors etc?

Tom, if you PM me a dropbox link I'll pop it on that server, to save you from being disabled again.

Also, the speed you're progressing on this is astonishing.

Edited September 4, 2010 by Josh04

[Guest hungary]

Nice work Tom!

If you release a new version with fixes, we only need:

Wifi bugs fix (signal strength and link speed are reported incorrectly)

Camera

[Guest AdamBaker]

Nice work Tom!

If you release a new version with fixes, we only need:

Wifi bugs fix (signal strength and link speed are reported incorrectly)

Camera

Could someone check if iwconfig reports signal strength and speed correctly, it is in /system/wifi but missing execute permission for some reason (at least it is on 1.5). That should give a clue as to where to look to fix it.

Does anyone know what is missing with the camera - presumable again there are some libs needed, http://pdk.android.com/online-pdk/guide/camera.html says we need a /system/lib/libcamera.so and it looks as though the Pulse version of that uses libmm-qcamera-tgt.so. I think libcameraservice.so is part of android and I'm not sure about libmmcamera.so. I haven't checked backward compatibility this time but it was clear effort had gone into it with the sensors so it likely has with the camera too.

[Guest rulerofkaos]

Does anyone know what is missing with the camera - presumable again there are some libs needed, http://pdk.android.com/online-pdk/guide/camera.html says we need a /system/lib/libcamera.so and it looks as though the Pulse version of that uses libmm-qcamera-tgt.so. I think libcameraservice.so is part of android and I'm not sure about libmmcamera.so. I haven't checked backward compatibility this time but it was clear effort had gone into it with the sensors so it likely has with the camera too.

As i understood the complete camera driver is missing.

[Guest flip360]

iwconfig gives me this output:

eth0 AR6000 802.11g

Bit Rate=54Mb/s

TxPower=15 dBm

Sensitivity=0/3

Retry:on

EncrytionKey=off

PowerManagement=off

LinkQuality: 35/94

SignalLevel: -60dBm

NoiseLevel: -96 dBm

Also i noticed that on every reboot the network switches automaticaly to 3G (after the fix), even if i checked "use 2G Networks Only"

Edited September 4, 2010 by flip360

[Guest rulerofkaos]

iwconfig gives me this output:

eth0 AR6000 802.11g

Bit Rate=54Mb/s

TxPower=15 dBm

Sensitivity=0/3

Retry:on

EncrytionKey=off

PowerManagement=off

LinkQuality: 35/94

SignalLevel: -60dBm

NoiseLevel: -96 dBm

Also i noticed that on every reboot the network switches automaticaly to 3G (after the fix), even if i checked "use 2G Networks Only"

This seems to be normal values for me, so why is it not working...

[Guest igor_anta]

iwconfig gives me this output:

eth0 AR6000 802.11g

Bit Rate=54Mb/s

TxPower=15 dBm

Sensitivity=0/3

Retry:on

EncrytionKey=off

PowerManagement=off

LinkQuality: 35/94

SignalLevel: -60dBm

NoiseLevel: -96 dBm

Also i noticed that on every reboot the network switches automaticaly to 3G (after the fix), even if i checked "use 2G Networks Only"

The switch to 3G on reboot happens on 2.1 as well, not all the time but it happens. So it's normal I guess.

[Guest Tom G]

Could someone check if iwconfig reports signal strength and speed correctly, it is in /system/wifi but missing execute permission for some reason (at least it is on 1.5). That should give a clue as to where to look to fix it.

Does anyone know what is missing with the camera - presumable again there are some libs needed, http://pdk.android.com/online-pdk/guide/camera.html says we need a /system/lib/libcamera.so and it looks as though the Pulse version of that uses libmm-qcamera-tgt.so. I think libcameraservice.so is part of android and I'm not sure about libmmcamera.so. I haven't checked backward compatibility this time but it was clear effort had gone into it with the sensors so it likely has with the camera too.

iwconfig will work correctly. The problem is that google added functions to wpa_supplicant and in those functions used none standard calls back to the driver. If the driver doesn't implement the calls (and it seems like most of them don't) then the wext driver from aosp doesn't work properly.

Take a look at section 8 of http://blog.linuxconsulting.ro/2010/04/por...to-android.html. All I have done is reset the error code to get it working. I haven't looked at really fixing it.

The 2.1 camera driver might work. If it is included in the build it fails because of a missing dependency, but it may just be that I need to include another file. I haven't looked at it much. There is source for a qcom camera driver on codeaurora (it might even be in aosp), but it is missing headers.

[Guest flip360]

any news on jit? is it stable? :P

[Guest shisan]

iwconfig will work correctly. The problem is that google added functions to wpa_supplicant and in those functions used none standard calls back to the driver. If the driver doesn't implement the calls (and it seems like most of them don't) then the wext driver from aosp doesn't work properly.

Take a look at section 8 of http://blog.linuxconsulting.ro/2010/04/por...to-android.html. All I have done is reset the error code to get it working. I haven't looked at really fixing it.

The 2.1 camera driver might work. If it is included in the build it fails because of a missing dependency, but it may just be that I need to include another file. I haven't looked at it much. There is source for a qcom camera driver on codeaurora (it might even be in aosp), but it is missing headers.

Well done，TOM！！

I would like to know when to release the next rom.

If the data will work?

Thank you guys

bb and you

Edited September 4, 2010 by shisan

[Guest McSpoon]

The 2.1 camera driver might work. If it is included in the build it fails because of a missing dependency, but it may just be that I need to include another file. I haven't looked at it much. There is source for a qcom camera driver on codeaurora (it might even be in aosp), but it is missing headers.

Not sure if it will help but you might have more luck with this:

http://gitorious.org/eclair-camera-drivers..._hardware_msm7k

It's been ages since I last looked at the code repositories but I recall the AOSP libcamera code being just a stub. And the codeaurora libcamera always seemed impossible to compile. Not sure if that has changed since but porting libcamera to the Pulse could be a pain. (plus I think Huawei may have hacked it to hell to add front-facing camera support.)