Error: TARGET_USES_UNCOMPRESSED_KERNEL is depreciated!

While starting my new attempt at an Oreo build for the H811/H815 (T-Mobile LG G4), I ran into an error, right off the bat:

Error: TARGET_USES_UNCOMPRESSED_KERNEL is depreciated!

I took a look at the make files that build the kernel. You can find them in the tasks folder under vendor/lineage/build/tasks/kernel.mk, which has a very lengthy explanation on how to name the kernel to build uncompressed images. Here is the pertinent part:

# BOARD_KERNEL_IMAGE_NAME = Built image name
# for ARM use: zImage
# for ARM64 use: Image.gz
# for uncompressed use: Image
# If using an appended DT, append ‘-dtb’
# to the end of the image name.
# For example, for ARM devices,
# use zImage-dtb instead of zImage.

So all I had to do was remove this line from my device/lge/g4-common/BoardConfigCommon.mk file:

TARGET_USES_UNCOMPRESSED_KERNEL := true

Because my kernel has the naming convention like so:

BOARD_KERNEL_IMAGE_NAME := Image

Which alludes to an uncompressed image already. I love it when the solution is easy!

Linux – keep it simple.

RWE: Ejabberd failed to start in Debian Wheezy

I had an old laptop that I decided would make a good toy for playing around with some lightweight server applications. It was a great candidate for several reasons. For one, I’m not using it for something else. Second, it was free. Third, it has a built in battery backup. Seemed like a win-win for me.

I’ve mentioned before that I am a member of SDF, the Super Dimensional Fortress (sdf.org), a free to the community Unix server that allows remote log in and numerous other goodies. Among those extras is the free jabber (XMPP) service for members. It’s really great, especially for free, but on several occaisions, it has been down for maintenance when I was waiting to receive a message to someone. That is a bit annoying. Obviously, server maintenance needs to happen, and I am really thankful to sdf.org for providing the service. I was just thinking that this would be a good excuse to try setting up my very own XMPP server.

Thus enter ejabberd. Ejabberd is an open source XMPP server that is available on most GNU/Linux distributions, including Debian Wheezy, which was already on my laptop. I looked up several tutorials on-line, and they all seemed to indicate that it would be super easy to set up, and I could get it up and running in a few minutes.

Screenshot_20171030-085330.png

Two days latter…..

I finally got it set up and running. No matter which tutorial I followed, none of them worked. It took hours of research (in between being a dad, husband, working person, etc.), but I finally figured out the problem. Not too many people are planning to set up their own XMPP server, and this problem is probably specific to Debian Wheezy, but I figured that I would show you the REAL steps to get this going.

The problem was that after installing ejabberd, even with the default settings, it would say:

starting service: ejabbered………………………………………………………..FAILED.

And there was no rhyme or reason why! I was only able to figure it out by running ejabberd live, so the output came to the terminal, and then I could see the issue.

# apt-get install ejabberd

If you are like me, it will fail after install.

# dpkg-reconfigure ejabberd

Set up your server name, admin, etc.

# ps axc|grep beam

# ps axc|grep epmd

Now kill both of those processes.

# nano /etc/default/ejabberd

Uncomment the ERLANG_NODE=ejabberd line, or add it if it does not exist. Save and exit that file. Not having this line caused my first failure.

Edit your  /etc/hosts file to look like this:
127.0.0.1 localhost

# The following lines are desirable for IPv6 capable hosts
::1 localhost ip6-localhost ip6-loopback
ff02::1 ip6-allnodes
ff02::2 ip6-allrouters

Essentially, you need to delete any “extra” hosts in the file, just leave the localhost ones. Later, more can be added to the file once ejabberd is running, but for some reason, having extra hosts in this file cause my second failure.

Now you can start ejabberd. I recommend that the first time, you start it “live” so you can see what is happening under the hood:

# /etc/init.d/ejabberd live

But in the future, you can just start it with the usual services tags:

# service ejabberd restart

You should now have a running instance of ejabberd! Of course, now you need to add a user, if you didn’t already. Here is an example:

# ejabberdctl register userName xmppServerAddress.net superSecretPassword

And there we go! At least for me, that was enough to set up a fully functional XMPP server with ejabberd on Debian Wheezy. Of course, this still needs tuning and enabling features like blocking, encryption, etc., but hey, at least now you have something to work with!

Linux – keep it simple.

error: undefined reference to ‘wifi_start_fstman’

Ah, the continuing saga of the far far away dream of building OmniRom Nougat for the LG G4. Yet again, another set of errors:

[CODE]
system/qcom/softap/sdk/qsap.c:705: error: undefined reference to ‘wifi_start_fstman’
system/qcom/softap/sdk/qsap.c:714: error: undefined reference to ‘wifi_stop_fstman’
system/qcom/softap/sdk/qsap.c:722: error: undefined reference to ‘is_fst_enabled’
clang++: error: linker command failed with exit code 1 (use -v to see invocation)
[/CODE]

So I took a look at the lines in question of system/qcom/softap/sdk/qsap.c:

[CODE]
int qsap_prepare_softap()
{
ALOGD(“Starting fstman\n”);
return wifi_start_fstman(TRUE);
}

int qsap_unprepare_softap()
{
ALOGD(“Stopping fstman\n”);
return wifi_stop_fstman(TRUE);
}

int qsap_is_fst_enabled()
{
return is_fst_enabled();
}
[/CODE]

No big deal, I thought. I could just wrap them in an #ifdef (if defined) statement, and only use them if they are actually defined. So I did.

[CODE]
int qsap_prepare_softap()
{
/* WJH */
#ifdef wifi_start_fstman
ALOGD(“Starting fstman\n”);
return wifi_start_fstman(TRUE);
#endif
}

int qsap_unprepare_softap()
{
/* WJH */
#ifdef wifi_stop_fstman
ALOGD(“Stopping fstman\n”);
return wifi_stop_fstman(TRUE);
#endif
}

int qsap_is_fst_enabled()
{
/* WJH */
#ifdef is_fst_enabled
return is_fst_enabled();
#endif
}
[/CODE]

You are all probably smarter than me, and see the error of my ways, but non the less, I ran it as so, and this was my next set of errors:

[CODE]
system/qcom/softap/sdk/qsap.c:707:1: error: control reaches end of non-void function [-Werror,-Wreturn-type]
}
^
system/qcom/softap/sdk/qsap.c:716:1: error: control reaches end of non-void function [-Werror,-Wreturn-type]
}
^
system/qcom/softap/sdk/qsap.c:724:1: error: control reaches end of non-void function [-Werror,-Wreturn-type]
}
^
[/CODE]

The problem was that the in int qsap_is_fst_enabled(), int qsap_unprepare_softap(), and int qsap_prepare_softap() integers all return nothing, or are “null”. That’s no good for follow on code, they have to have some value. So, I thought about it for a while, and right or wrong, I decided to use an #ifndef (if not defined) flag.
If the functions were defined, they would be used, if they are not defined, then I gave a return value, arbitrary as they may be, but what I suspect is the “right answer”. Take a look:

[CODE]
int qsap_prepare_softap()
{
/* WJH */
#ifdef wifi_start_fstman
ALOGD(“Starting fstman\n”);
return wifi_start_fstman(TRUE);
#endif
#ifndef wifi_start_fstman
ALOGD(“Starting fstman\n”);
return 1;
#endif
}

int qsap_unprepare_softap()
{
/* WJH */
#ifdef wifi_stop_fstman
ALOGD(“Stopping fstman\n”);
return wifi_stop_fstman(TRUE);
#endif
#ifndef wifi_stop_fstman
ALOGD(“Stopping fstman\n”);
return 1;
#endif
}

int qsap_is_fst_enabled()
{
/* WJH */
#ifdef is_fst_enabled
return is_fst_enabled();
#endif
#ifndef is_fst_enabled
return 0;
#endif
}
[/CODE]

And viola! This time it worked. What result that will have in the end? I’m not sure. Hopefully it still functions properly, but that remains to be seen….

Linux – keep it simple.

 

Function “appears in overlay but not in the base package”

Ah, OmniRom. Yes, I am still working on it. Hopefully you are not getting tired of these errors, but I find it helpful to log them here, and hopefully others will also. Previously, we looked at ways to use <add-resource> to add functions into a rom, but here we are just going to disable an overlay that we can work on later.

[CODE]
FAILED: /bin/bash -c “(touch out/target/common/obj/APPS/framework-res_intermediates/zipdummy ) && ((cd out/target/common/obj/APPS/framework-res_intermediates/ && jar cf package-export.apk zipdummy) ) && (zip -qd out/target/common/obj/APPS/framework-res_intermediates/package-export.apk zipdummy ) && (rm out/target/common/obj/APPS/framework-res_intermediates/zipdummy ) && (out/host/linux-x86/bin/aapt package -u -x –private-symbols com.android.internal -z –pseudo-localize -M frameworks/base/core/res/AndroidManifest.xml -S device/lge/g4-common/overlay/frameworks/base/core/res/res -S frameworks/base/core/res/res -A frameworks/base/core/res/assets –min-sdk-version 25 –target-sdk-version 25 –product default –version-code 25 –version-name 7.1.2 –skip-symbols-without-default-localization -F out/target/common/obj/APPS/framework-res_intermediates/package-export.apk )”
device/lge/g4-common/overlay/frameworks/base/core/res/res/values/config.xml:277: error: Resource at config_supportDoubleTapWake appears in overlay but not in the base package; use <add-resource> to add.
[/CODE]

So, as you can see, we have something in our overlay that is supposed to lay over the top of some source code, but the base source code doesn’t have that function to be “overlayed”. Here is where we see the overlay in device/lge/g4-common/overlay/frameworks/base/core/res/res/values/config.xml:277:

[CODE]
<!– Whether device supports double tap to wake –>
<bool name=”config_supportDoubleTapWake”>true</bool>
[/CODE]

So, we changed it to:

[CODE]
<!– Whether device supports double tap to wake
<bool name=”config_supportDoubleTapWake”>true</bool> WJH –>
[/CODE]

By commenting it out, it will no longer “overlay” it onto something that doesn’t exist. Be sure to take a look at my other article for how to use <add resource> to actually add an overlay item to the base package.

Linux – keep it simple.

 

error: use of undeclared identifier ‘LED_ADJUSTMENT_G’

After synching my source with AOKP for Nougat 7.1.2, I ran into this error when trying to build for the JFLTETMO:

[CODE]
target thumb C: lights.MSM8960 <= hardware/samsung/liblights/lights.c
hardware/samsung/liblights/lights.c:215:40: error: use of undeclared identifier ‘LED_ADJUSTMENT_R’
int red = ((color >> 16) & 0xFF) * LED_ADJUSTMENT_R;
^
hardware/samsung/liblights/lights.c:216:41: error: use of undeclared identifier ‘LED_ADJUSTMENT_G’
int green = ((color >> 8) & 0xFF) * LED_ADJUSTMENT_G;
^
hardware/samsung/liblights/lights.c:217:33: error: use of undeclared identifier ‘LED_ADJUSTMENT_B’
int blue = (color & 0xFF) * LED_ADJUSTMENT_B;
^
hardware/samsung/liblights/lights.c:258:31: error: use of undeclared identifier ‘LED_BRIGHTNESS_BATTERY’
adjusted_brightness = LED_BRIGHTNESS_BATTERY;
^
hardware/samsung/liblights/lights.c:261:31: error: use of undeclared identifier ‘LED_BRIGHTNESS_NOTIFICATION’
adjusted_brightness = LED_BRIGHTNESS_NOTIFICATION;
^
hardware/samsung/liblights/lights.c:264:31: error: use of undeclared identifier ‘LED_BRIGHTNESS_ATTENTION’
adjusted_brightness = LED_BRIGHTNESS_ATTENTION;
^
6 errors generated.
[/CODE]

Pretty simple, as we have talked about this before. In the past, I actually removed these lines from the light.c file, but this time I decided I would leave them in, and declare them instead. Essentially, the outcome is the same though for this build.

What these lines do is allow for custom liblight configurations. Perhaps a particular phone or tablet has a built in LED notification light where the green is brighter than the red or blue. In that case, every mixed color made from those LEDs will look strange. This gives the rom maker the option to “dim” down one or all three of the lights. The same applies for the brightness of the battery, notification, or attention lights.

As you can see below, I just declared them with the default values:

[CODE]
return err;
}

// WJH declaring these undeclared lights.
int LED_ADJUSTMENT_R = 1;
int LED_ADJUSTMENT_G = 1;
int LED_ADJUSTMENT_B = 1;
// WJH declaring these undeclared lights.

static int calibrate_color(int color, int brightness)
{
int red = ((color >> 16) & 0xFF) * LED_ADJUSTMENT_R;
int green = ((color >> 8) & 0xFF) * LED_ADJUSTMENT_G;
int blue = (color & 0xFF) * LED_ADJUSTMENT_B;

return (((red * brightness) / 255) << 16) + (((green * brightness) / 255) << 8) + ((blue * brightness) / 255);
}

…………EDITED FOR SPACE………………

led->delay_off = state->flashOffMS;
break;
default:
return -EINVAL;
}

// WJH declaring these brightness adjustments.
int LED_BRIGHTNESS_BATTERY = 255;
int LED_BRIGHTNESS_NOTIFICATION = 255;
int LED_BRIGHTNESS_ATTENTION = 255;
// WJH declaring these brightness adjustments.

switch (type) {
case TYPE_BATTERY:
adjusted_brightness = LED_BRIGHTNESS_BATTERY;
break;
case TYPE_NOTIFICATION:
adjusted_brightness = LED_BRIGHTNESS_NOTIFICATION;
break;
case TYPE_ATTENTION:
adjusted_brightness = LED_BRIGHTNESS_ATTENTION;
break;
default:
adjusted_brightness = 255;
}
[/CODE]

 

For the LED_ADJUSTMENT_R/G/B, I gave a value of 1. If you look at the math, it is used in the next formulae as (a & b) * LED_ADJUSTMENT_R/G/B. So if I set a value of 0, then the mathmatical outcome will be 0. If I set it at 1, then it will always be (a & b).

For the LED_BRIGHTNESS_BATTERY and so forth, I used the adjusted_brightness of 255, which is the default.

For both of these sets, I could have just removed them all together. But I build for several phones using the same source. By leaving them in, I should have an error if the device tree declared what is already declared in the lights.c file. Or at least that is my hope. In that case, then I could remove my handy work to make room for the actual values.

Linux – keep it simple.

Fixing the blank developer settings on SlimRoms Nougat

After building SlimRoms Nougat for the H811 (LG G4 T-Mobile), I ran into an odd artifact: when you enabled and opened the developer settings, it would be blank, or crash. I had absolutely no idea why that would be, nor where to begin looking into a fix.

So, I hopped over to SlimRoms’ gerrit: https://review.slimroms.org/#/q/status:merged

Once there, I looked at merged (accepted or fixed) labels, and typed “developer” in the search box. A lot of things came up, but there was a noticeable trend among the contenders, a crash or ANR report when opening the developer settings.

In every case they used the same fix, so I decided to give it a try. It looks like the common factor in every case was the FRP, or factory reset protection in the system.prop files for the devices. I took a look at mine:

# Factory Reset Protection
PRODUCT_PROPERTY_OVERRIDES += \
ro.frp.pst=/dev/block/bootdevice/by-name/persistent

and edited them as such (just commented out) :

# Factory Reset Protection
# PRODUCT_PROPERTY_OVERRIDES += \
# ro.frp.pst=/dev/block/bootdevice/by-name/persistent
# WJH commented out as this prevents the developer options from opening.

And it was fixed! It’s great when the fix is so easy!

Linux – keep it simple!

undeclared identifier ‘POWER_HINT_LAUNCH_BOOST’

So, a new one for me. I’m building a rom for the H811, the T-Mobile variant of the LG G4. Currently building the tried and true AOKP 7.1.2 that I have used for my other phones, and I ran into an error:

[CODE]
target thumb C: power.msm8992_32 <= device/qcom/common/power/power-8992.c
device/qcom/common/power/power-8992.c:211:17: error: use of undeclared identifier 'POWER_HINT_LAUNCH_BOOST'; did you mean 'POWER_HINT_CPU_BOOST'?
if (hint == POWER_HINT_LAUNCH_BOOST) {
^~~~~~~~~~~~~~~~~~~~~~~
POWER_HINT_CPU_BOOST
hardware/libhardware/include/hardware/power.h:71:5: note: 'POWER_HINT_CPU_BOOST' declared here
POWER_HINT_CPU_BOOST = 0x00000110,
^
1 error generated.
[/CODE]

The suggested option was to change POWER_HINT_LAUNCH_BOOST to POWER_HINT_CPU_BOOST, but I didn't like that, since just a dozen lines later in the code, there was already an if/then statement for POWER_HINT_CPU_BOOST. Esentially, we need to either declare POWER_HINT_LAUNCH_BOOST, or get rid of it in our code. In this case, I decided to scrap it because if it was needed, then it would have existed already. Strange logic perhaps, but what we see is that if it is not called, it will not return anything anyways. If it is called, it just supplies some information that I think we can live without.

[/CODE]
// WJH was POWER_HINT_LAUNCH_BOOST
// WJH if (hint == POWER_HINT_LAUNCH_BOOST) {
// WJH launch_boost_info_t *info = (launch_boost_info_t *)data;
// WJH if (info == NULL) {
// WJH ALOGE("Invalid argument for launch boost");
// WJH return HINT_HANDLED;
// WJH }

// WJH ALOGV("LAUNCH_BOOST: %s (pid=%d)", info->packageName, info->pid);

// WJH int duration = 2000;
// WJH int resources[] = { SCHED_BOOST_ON, 0x20C };

// WJH start_prefetch(info->pid, info->packageName);
// WJH interaction(duration, ARRAY_SIZE(resources), resources);

// WJH return HINT_HANDLED;
// WJH }

if (hint == POWER_HINT_CPU_BOOST) {
int duration = *(int32_t *)data / 1000;
int resources[] = { SCHED_BOOST_ON };
………….EDITED FOR SPACE…………….
[/CODE]

That will cut it out of the loop, which means the loss of a feature, but I think it is one we can live without.

Linux – keep it simple.

POWER_HINT_LAUNCH_BOOST is not declared in this scope!

While building AOKP 7.1.2 for the kltevwz (the Verizon Samsung Galaxy S5), I ran into this error:

[CODE]

device/qcom/common/power/power-8974.c:143:17: error: ‘POWER_HINT_LAUNCH_BOOST’ was not declared in this scope …….Edited for space…….. make: *** Waiting for unfinished jobs….

[/CODE]

So, I headed over to device/qcom/common/power/power-8974.c, and this is what I saw:

[CODE]

if (hint == POWER_HINT_LAUNCH_BOOST) {
int duration = 2000;
int resources[] = { CPUS_ONLINE_MIN_3,
CPU0_MIN_FREQ_TURBO_MAX, CPU1_MIN_FREQ_TURBO_MAX,
CPU2_MIN_FREQ_TURBO_MAX, CPU3_MIN_FREQ_TURBO_MAX };

interaction(duration, ARRAY_SIZE(resources), resources);

return HINT_HANDLED;
}

[/CODE]

Looked pretty normal to me. However, when comparing to LineageOS’s 14.1 files, I saw that they dropped the “boost” from the name of the hint. So I edited mine to match, like so:

[CODE]

if (hint == POWER_HINT_LAUNCH) {
int duration = 2000;
int resources[] = { CPUS_ONLINE_MIN_3,
CPU0_MIN_FREQ_TURBO_MAX, CPU1_MIN_FREQ_TURBO_MAX,
CPU2_MIN_FREQ_TURBO_MAX, CPU3_MIN_FREQ_TURBO_MAX };

interaction(duration, ARRAY_SIZE(resources), resources);

return HINT_HANDLED;
}

[/CODE]

And, it built that part successfully! I like the easy problems!

Linux – keep it simple.

Yet again, error: package com.android.okhttp does not exist

By now, I think I would remember this one by heart, yet every time, this error throws me for a loop….

[CODE]
external/apache-http/../../frameworks/base/core/java/android/net/http/SslCertificate.java:19: error: cannot find symbol
import com.android.internal.util.HexDump;
^
symbol: class HexDump
location: package com.android.internal.util
external/apache-http/../../frameworks/base/core/java/android/net/http/SslCertificate.java:42: error: package com.android.org.bouncycastle.asn1.x509 does not exist
import com.android.org.bouncycastle.asn1.x509.X509Name;
^
external/apache-http/../../frameworks/base/core/java/android/net/http/HttpResponseCache.java:19: error: package com.android.okhttp does not exist
import com.android.okhttp.Cache;
^
external/apache-http/../../frameworks/base/core/java/android/net/http/HttpResponseCache.java:20: error: package com.android.okhttp does not exist
import com.android.okhttp.AndroidShimResponseCache;
^
external/apache-http/../../frameworks/base/core/java/android/net/http/HttpResponseCache.java:21: error: package com.android.okhttp does not exist
import com.android.okhttp.OkCacheContainer;
^
external/apache-http/../../frameworks/base/core/java/android/net/http/HttpResponseCache.java:152: error: cannot find symbol
public final class HttpResponseCache extends ResponseCache implements Closeable, OkCacheContainer {/home/alaskalinuxuser/compile/20170314_AOKP7/prebuilts/tools/common/m2/repository/com/squareup/okhttp/okhttp/2.5.0
^
symbol: class OkCacheContainer
external/apache-http/../../frameworks/base/core/java/android/net/http/HttpResponseCache.java:154: error: cannot find symbol
private final AndroidShimResponseCache delegate;
^
symbol: class AndroidShimResponseCache
location: class HttpResponseCache
external/apache-http/../../frameworks/base/core/java/android/net/http/HttpResponseCache.java:156: error: cannot find symbol
private HttpResponseCache(AndroidShimResponseCache delegate) {
^
symbol: class AndroidShimResponseCache
location: class HttpResponseCache
external/apache-http/../../frameworks/base/core/java/android/net/http/HttpResponseCache.java:296: error: cannot find symbol
public Cache getCache() {
^
symbol: class Cache
location: class HttpResponseCache
[/CODE]

And, from the same website that I fail to bookmark every time:

https://issuetracker.google.com/issues/37130763

These are caused by missing some package dependency. Following patch is used to fix it (for external/apache-http/Android.mk):

— a/Android.mk
+++ b/Android.mk
@@ -95,7 +95,7 @@ LOCAL_SRC_FILES += \
../../frameworks/base/core/java/android/net/http/SslError.java \

-LOCAL_JAVA_LIBRARIES := $(apache_http_java_libs)
+LOCAL_JAVA_LIBRARIES := $(apache_http_java_libs) okhttp bouncycastle framework
LOCAL_MODULE_CLASS := JAVA_LIBRARIES
LOCAL_DROIDDOC_SOURCE_PATH := $(LOCAL_PATH)/src \
$(LOCAL_PATH)/android \

That’s it. Those few ines make the whole thing work. Hopefully, I’m saving someone else from scouring the internet for this problem. Again.

Linux – keep it simple.

SyntaxError: invalid syntax

As I was wrapping up my work on the upgrade to 7.1.2 on the S4, I ran into this error, which prevented the compiler from building the userdata.img file. Here’s the output:

[CODE]
get/product/jfltetmo/userdata.img
File “./build/tools/releasetools/build_image.py”, line 157
print cmd
^
SyntaxError: invalid syntax
[/CODE]

The fix was actually quite easy. I needed to encapsulate the cmd in parenthesis, like this:

./build/tools/releasetools/build_image.py

[CODE]
cmd = “fec -e -p %d %s %s %s” % (padding_size, sparse_image_path,
verity_path, verity_fec_path)
print (cmd) # WJH
status, output = commands.getstatusoutput(cmd)
[/CODE]

The proof is in the pudding, but what help was how the cmd is defined, then notice that the “status, output” line also puts cmd in parenthesis. So, I did the same for the print cmd line, and it worked!

Linux – keep it simple.