git-master on ADP1/dream (third take)

[Jean-Baptiste Queru]

I've had to revert some of the changes that I had done (they didn't
fit in the "big picture", and I don't have enough time to make them
fit well).

Current state of git-master of ADP1/dream as I can see it:

-works after latest cupcake merge (which was done yesterday night)
-needs change 7699 manually applied to get wifi.
-needs change 7345 manually applied to get the same APNs preconfigured
as in the ADP1.
-need to tweak development/apps/SdkSetup/Android.mk to use
LOCAL_MODULE_TAGS := eng development and
development/apps/SdkSetup/AndroidManifest.xml to add the
WRITE_SETTINGS in addition to WRITE_SECURE_SETTINGS in order for the
device to be provisioned.
-bluetooth, GPS, compass/accelerometer, and HW OpenGL are broken.
-none of the Google apps (mail, talk, market, maps...)

JBQ

--
Jean-Baptiste M. "JBQ" Queru
Android Engineer, Google.

[Aaron]

Good work JBQ. Still a lot of things are needed to be functional.

I flashed cupcake onto my phone a few days ago just to try it out. I
know it's still a development build but everything was very laggy. I
am hoping my phone just has other incompatibilities and it's not the
OS recessing in speed.

[oth...@googlemail.com]

JBQ, you are saving people like me many hours of trying to fix
something that does not yet work. I cannot say how much I appreciate
to find this info here. While not ready for the public yet, I must say
I like 1.5 (and Android as such) better every day. WiFi does indeed
work well now.
- However, when I replace the prebuild kernel with one I build from
the master kernel folder, WiFi does again not work.
- There is also this issue with the onscreen keyboard not advancing
the cursor.
- And I am getting "couldn't load <libhgl.so> library" messages and
wonder if maybe there is an easy fix for the OpenGL issue in cupcake?
I don't see this msg when I run 1.0, where OpenGL works fine.
Tim

[Jean-Baptiste Queru]

-The on-screen keyboard is a new feature and is still work in
progress, so seeing bugs there is to be expected.

-I have no visibility over the process to re-build your own kernel and
wlan.ko module (I know that you have to rebuild them together or wifi
won't work, but I don't know whether that's the issue you're running
into). The android-kernel mailing list would be more appropriate for
such questions.

-Hardware OpenGL is not currently working in the open-source tree,
most probably because it's missing some proprietary binaries that
aren't available through git. There might be an easy fix, but I know
so little about OpenGL that I can't help here.

JBQ

--

[Disconnect]

Just for the record, and for those not on IRC, I do builds 1-2 times a day (as changes come in) and post them to andblogs.net/images.

More on-topic, I know that right now wlan.ko and kernel are both prebuilt. And at least when cupcake was new, San just did manual builds (inside the source tree, outside the build system.) and chucked them in prebuilt.

[Joel Knighton]

Disconnect, as long as you're posting about your images, I have a few questions about them and the latest builds of Cupcake.  With the most recent image on your site, I can only boot it if I remove GoogleContactProvider.apk (or something similar, can't quite recall).  From that point, the home and call buttons do not work, and the phone does not receive incoming calls.  Is this typical for the cupcake branch?

--
Joel Knighton

[Joel Knighton]

Okay, Logcat says cupcake is rejecting incoming calls because it is not provisioned.  I have added the necessary APNs.  What else is required?

--
Joel Knighton

[Jean-Baptiste Queru]

you need to tweak development/apps/SdkSetup/Android.mk to use
LOCAL_MODULE_TAGS := eng development

JBQ

--

[Joel Knighton]

Oh, okay.  I would suggest Disconnect do that on his latest builds, then.  ;)  I'm off to build, I guess.

--
Joel Knighton

[Disconnect]

Hmm. I had that :( but not currently.

Fixing it now...

[Disconnect]

OK I had the other eng development change. :/

So .. should it be a policy that a change that breaks the aosp build goes in with fixes? (Or at least a warning.) "Breaks" including things like "makes it depend {or depend again} on proprietary bins"..

Not picking on jbq, he's just in the front cuz of all the work he's doing :)

[Jean-Baptiste Queru]

That'd be cool, yeah. In this case it's a tiny bit different - it was
broken to start with, I attempted to fix it, and I had to revert my
fix, so we're back to square one. That puts us in a deadlocked
situation: I can't make the system go "back" to a state that's both
acceptable and working, and I don't have a fix to go "forward" into a
state that's working. We're gonna have to live with this until someone
steps in and finds the time to make an acceptable fix.

JBQ

[Disconnect]

Not an unreasonable position, although maybe the android.mk change should be in by default? (I'll submit it if you want, along with the other minor changes.)

Any other 'gotchas' that aren't on the building-for-dream page? So far we have the (optional) PRODUCT-htc_dream-user make target and the 2 Android.mk changes..

[Jean-Baptiste Queru]

Changing Android.mk won't quite fly either, because that'll get in the
way of Google's internal builds that do include the proprietary
SetupWizard.

It seems to be that an answer along those lines would be to make that
configurable at the product level, so that the builds from the
open-source tree can include SdkSetup while builds done by Google (or
Google's partners who have SetupWizard) don't.

JBQ

[Disconnect]

So the proprietary builds are going to take precedence over the AOSP tree?

Thats a very dark road....

[Jean-Baptiste Queru]

No, we just need to find a solution that satisfies the needs of both
the builds that do and the builds that do not have a proprietary
SeupWizard (whether that's Google's or any other one).

Breaking one side to fix the other just shifts the problem, it doesn't fix it.

JBQ

[Disconnect]

If the goal is an open AOSP, then the side adding proprietary pieces needs to be treated as a closed-source fork. Not the other way around...

So my vote (fwiw) is that the fixes go into the tree, and the "we need to do something different" framework is built to support the closed-source forks, rather than pushing it into the open tree and saying "thats too bad, but we have a closed fork that is more important."

[Jean-Baptiste Queru]

-The goal of the AOSP is to provide a software platform for
OEMs/ODMs/integrators to build phones (and other devices) on top of.
Changing the code in a way that would make that harder is a step in
the wrong direction.

-On top of that, you want Google to be able to move its
Android-related development activities in the open. I want it too.
Google wants it too. The entire Android community wants it too. Let's
not start to make code changes which we know will make that transition
harder than it already is.

There've been some discussions already on android-framework about a
solution to the specific problem at hand that would be both effective
and acceptable. I guess it'd take 1 to 2 man-days of work to implement
it, and I don't have that much time to spare.

We're on a slippery slope that leads straight into the bottomless pit
of pointless Internet arguments. Feel free to continue the discussion
on android-discuss...

JBQ

[Disconnect]

Oooh, a straw-man followed by a godwin. Nice. For the record, you can pull me out of "we" in "make that harder", "we know will make that transition harder", etc. I don't know these things, because I don't have access to your closed-source forks. I just know that the open source tree is getting incompatible checkins - this one in particular is a bad example of that, admittedly - due to those closed-source products, and the policy is either going to be "too bad, closed trees are the customer, aosp is the afterthought" or "aosp is the core, closed trees are the fork."

Feel free to ignore this. (And in other news, thanks for the merge logs. I'll post them when i finish my meetings today..)

[Jean-Baptiste Queru]

The goal is to have a single tree, in which Googlers working on AOSP
contribute directly, so that the discussion about which tree is
upstream and which tree is fork doesn't need to happen. Let's make
that happen as quickly as possible.

Yeah, "we" was quite a stretch in my previous post. Not everyone has
spent hours on this specific issue in the last week or so. Sorry about
that.

JBQ

[Jeff Hamilton]

I haven't looked into this issue at all, but I'm wondering if we can
solve it by using the same build mechanism GoogleContactsProvider uses
to replace ContactsProvider:

LOCAL_OVERRIDES_PACKAGES := ContactsProvider

That instructs the build system to explicitly exclude ContactsProvider
if GoogleContactsProvider is included in the packaging for the current
build.

-Jeff

[Jean-Baptiste Queru]

If I understand correctly, the idea would be to include SdkSetup in
"all" builds (with a different name, of course), and to use
LOCAL_OVERRIDE_PACKAGES in Google's SetupWizard's makefile to remove
SdkSetup.

Yes, that would work fine.

How should SdkSetup be named, and where should it be moved? I don't
think it belongs in development, but I also don't think that it's
anywhere near big enough to justify its own git repository. Should it
just go straight into framework/policies/base?

JBQ

[Dianne Hackborn]

I think it belongs in development.  It is really a part of configuring the SDK environment, which is pretty strongly in the realm of development.  As a quick way to get this working for people, doing this certainly make sense, but you wouldn't ship a device with SdkSetup on it.

--
Dianne Hackborn
Android framework engineer
hac...@android.com

Note: please don't send private questions to me, as I don't have time to provide private support.  All such questions should be posted on public forums, where I and others can see and answer them.

[Jean-Baptiste Queru]

How then would you ship a device without any setup wizard? The only
relevant thing that SdkSetup does at this point is to set the
"provisioned" bit. All the other aspects are/were related to the fact
that SDK 1.0r1 didn't contain a settings app, which isn't an issue any
more (and therefore those lines of code should be removed).

JBQ

[oth...@googlemail.com]

It is not possible to build for dream for quite a while... Just
downloaded huge updates from master and tried again (with 7699, 7345
and everything else). Today's build error: no rule to build target
»out/target/product/dream/obj/lib/libopencoreplayer.so«

[JJ]

That is because two other required changes are not merged( there are
some issues with submission process).
http://review.source.android.com/Gerrit#change,8298
http://review.source.android.com/Gerrit#change,8299
You can download these two and give it a try.

-JJ

On Jan 26, 2:49 pm, "oth...@googlemail.com" <oth...@googlemail.com>
wrote:

[Aaron]

JBQ, are there any updates on bluetooth, openGL, or any of the other
apps (Market, IM, and etc) working on ADP1?

Are there anyone actively looking at these issues or will it be
dropped and forgotten?

[Jean-Baptiste Queru]

Both those issues (drivers and apps) are important and discussed on a
regular basis, and they're not forgotten.

JBQ

--

[oth...@googlemail.com]

> That is because two other required changes are not merged( there are
> some issues with submission process).
>            http://review.source.android.com/Gerrit#change,8298
>            http://review.source.android.com/Gerrit#change,8299
> You can download these two and give it a try.

This let's me build fine. WiFi is working. However, signal appears to
be week and is getting lost rather quickly.
Well, I got two questions: is the cupcake/dream kernel supposed to
support ext2 now? I read this somewhere.
And is there really nobody out there with *any* idea on how to get HW
accelerated graphics working?
Tim

[Jean-Baptiste Queru]

-I think that the kernel is now built with e2fs/e3fs support indeed.

-The problem for HW graphics is to find someone who has both the
skills and the time to try to fix it. I surely know someone who has
the skills but not the time.

JBQ

--

[oth...@googlemail.com]

I don't know. Somehow I think it shouldn't take too much time for a
skilled person to fix the HW graphics issue. It seems so lightly
broken.
T.

[Jean-Baptiste Queru]

That doesn't mean that it's any of those people's top priority.

JBQ

--

[oth...@googlemail.com]

> That doesn't mean that it's any of those people's top priority.

Can you please forward the issue? I can offer an OpenGL related fix in
return.

In v1.0, where HW acceleration is available, every use of the GL API's
results in objects being generated and released to garbage from within
luni/platform/PlatformAddressFactory. In my GL animations, this
results in very visible GC hiccups every couple of seconds. I would be
happy to contribute a simple 20 line patch that fixes this issue,
should this be of any use. But unfortunately, I am not in a position
to fix the driver issue myself.

T.

[Jean-Baptiste Queru]

I have already forwarded the issue a while back, and I've recently
confirmed that the people who need to know about it do know about it.

JBQ

--

[Aaron]

I hope project leads will push this and make it a higher priority.
With the right experts in hand, making cupcake work on ADP1 shouldn't
take a significant portion of time (possibly 1-2 days). This is
better than pushing off the issue and making the whole community wait
endlessly when we can be providing help in terms of testing and
development.

[Dianne Hackborn]

Hi Aaron,

I definitely understand the desire for this and would really like to make this all work as soon as possible.  However, there are various other factors involved that hopefully can help explain the current situation.

The project lead who would be making such things a high priority is often the same person who would be doing the work. :)  The Android team is relatively small, and we all work with many competing demands on our time.

In addition, the schedule for Cupcake is extremely aggressive, with some very hard deadlines for delivery.  Having an engineer take 1-2 days on this right now essentially means dropping a feature from Cupcake.  For the most part, that is difficult to justify at this point.  For example, the person responsible for OpenGL could either spend time right now on making the open-source Cupcake source work with acceleration on the G1, or could add an animation when changing orientations that I have requested.  Right now the latter is I think much more important (and still something that is tough to fit in the schedule), so it just makes sense for work on the open-source tree to take a back seat for a bit.

Because of the tight schedule for Cupcake, we are also taking very few contributions for it from the open source tree (to avoid destabilizing things more than needed), and are still working in the same internal source control system that the 1.0 release was made from.  This unfortunately means that most of the primary engineers are working in a different environment from the source tree, which is going to cause instability there.  It's very unfortunate, but a necessity due to the schedule.

So having the open-source Android code in a state where you can easily build and install it on a G1 is most certainly a goal, but due to other priorities not something we focus completely on at this point.  We hope to move much more of the existing development process out to the open-source tree after Cupcake, and in conjunction with that will continue to address these issues.  It's just not going to happen immediately, but definitely will happen, and I hope I can ask for a little patience with us.

Work is being done every day on building out the open source environment (gerrit2 was a big improvement in that direction), but it's going to take a little while for it to all get done.

[Aaron]

Thanks for the clarification Dianne. I agree with everything you
said. I apologize for all the whining and the tone of the original
message.

> hack...@android.com

[Mike Lockwood]

FYI - Dream support for GPS and bluetooth is now working in
git-master, using the extract-files.sh approach described at
http://source.android.com/documentation/building-for-dream. The
bluetooth support includes work in progress on A2DP and AVRCP. And I
have a change that fixes hardware OpenGL in the queue:
http://review.source.android.com/8743

Note: the changes that fix GPS and bluetooth require a clean build to work.

Mike

On Wed, Jan 21, 2009 at 7:45 PM, Jean-Baptiste Queru <j...@google.com> wrote:
>
> -bluetooth, GPS, compass/accelerometer, and HW OpenGL are broken.

--
Mike Lockwood
Google android team

[Micael]

> FYI - Dream support for GPS and bluetooth is now working in

> git-master, using the extract-files.sh approach described athttp://source.android.com/documentation/building-for-dream.  The

> bluetooth support includes work in progress on A2DP and AVRCP.  And I
> have a change that fixes hardware OpenGL in the queue:http://review.source.android.com/8743

I tried yesterday to include that library, but it does not seem to be
enough..
With the libhgl.so library added it instead complains about some
missing files in /system/lib/hw/.....trout.so
However the "hw" directory don't exist on my ADP1 image.

But it's a step forward!

/Micael

[Giam Teck Choon]

On Feb 2, 12:19 pm, Mike Lockwood <lockw...@android.com> wrote:
> FYI - Dream support for GPS and bluetooth is now working in

> git-master, using the extract-files.sh approach described athttp://source.android.com/documentation/building-for-dream.  The

> bluetooth support includes work in progress on A2DP and AVRCP.  And I
> have a change that fixes hardware OpenGL in the queue:http://review.source.android.com/8743
>
> Note: the changes that fix GPS and bluetooth require a clean build to work.

I created a new folder and did the repo sync to make sure it is
clean. The build not even complete with the following error as I run
make showcommands after make clean && make dataclean && make
installclean to be sure it is clean :p


target StaticLib: libSR_AudioIn (out/target/product/dream/obj/
STATIC_LIBRARIES/libSR_AudioIn_intermediates/libSR_AudioIn.a)
prebuilt/linux-x86/toolchain/arm-eabi-4.2.1/bin/arm-eabi-ar crs out/
target/product/dream/obj/STATIC_LIBRARIES/libSR_AudioIn_intermediates/
libSR_AudioIn.a out/target/product/dream/obj/STATIC_LIBRARIES/
libSR_AudioIn_intermediates/audioinwrapper.o out/target/product/dream/
obj/STATIC_LIBRARIES/libSR_AudioIn_intermediates/audioin.o out/target/
product/dream/obj/STATIC_LIBRARIES/libSR_AudioIn_intermediates/
filter.o
prebuilt/linux-x86/toolchain/arm-eabi-4.2.1/bin/arm-eabi-gcc -mthumb-
interwork -o out/target/product/dream/obj/lib/crtend_android.o -c
bionic/libc/arch-arm/bionic/crtend.S
target Executable: AudioHardwareRecordLoop (out/target/product/dream/
obj/EXECUTABLES/AudioHardwareRecordLoop_intermediates/LINKED/
AudioHardwareRecordLoop)
prebuilt/linux-x86/toolchain/arm-eabi-4.2.1/bin/arm-eabi-g++ -nostdlib
-Bdynamic -Wl,-T,build/core/armelf.x -Wl,-dynamic-linker,/system/bin/
linker -Wl,--gc-sections -Wl,-z,nocopyreloc -o out/target/product/
dream/obj/EXECUTABLES/AudioHardwareRecordLoop_intermediates/LINKED/
AudioHardwareRecordLoop -Lout/target/product/dream/obj/lib -Wl,-rpath-
link=out/target/product/dream/obj/lib -lsrec_jni -llog -lutils -
lhardware_legacy -lc -lstdc++ -lm out/target/product/dream/obj/lib/
crtbegin_dynamic.o out/target/product/dream/obj/EXECUTABLES/
AudioHardwareRecordLoop_intermediates/src/
AudioHardwareRecordLoop.o out/target/product/dream/obj/
STATIC_LIBRARIES/libSR_AudioIn_intermediates/libSR_AudioIn.a -Wl,--no-
undefined /home/choon/droid/prebuilt/linux-x86/toolchain/arm-
eabi-4.2.1/bin/../lib/gcc/arm-eabi/4.2.1/interwork/libgcc.a out/target/
product/dream/obj/lib/crtend_android.o
/home/choon/droid/prebuilt/linux-x86/toolchain/arm-eabi-4.2.1/bin/../
lib/gcc/arm-eabi/4.2.1/../../../../arm-eabi/bin/ld: warning:
librpc.so, needed by out/target/product/dream/obj/lib/libgps.so, not
found (try using -rpath or -rpath-link)
out/target/product/dream/obj/lib/libgps.so: undefined reference to
`xprt_register'
out/target/product/dream/obj/lib/libgps.so: undefined reference to
`svc_freeargs'
out/target/product/dream/obj/lib/libgps.so: undefined reference to
`svc_sendreply'
out/target/product/dream/obj/lib/libgps.so: undefined reference to
`svcerr_decode'
out/target/product/dream/obj/lib/libgps.so: undefined reference to
`xdr_bytes'
out/target/product/dream/obj/lib/libgps.so: undefined reference to
`clnt_create'
out/target/product/dream/obj/lib/libgps.so: undefined reference to
`xdr_void'
out/target/product/dream/obj/lib/libgps.so: undefined reference to
`xdr_quad_t'
out/target/product/dream/obj/lib/libgps.so: undefined reference to
`svcerr_noproc'
out/target/product/dream/obj/lib/libgps.so: undefined reference to
`xdr_vector'
out/target/product/dream/obj/lib/libgps.so: undefined reference to
`svcrtr_create'
out/target/product/dream/obj/lib/libgps.so: undefined reference to
`svc_getargs'
out/target/product/dream/obj/lib/libgps.so: undefined reference to
`xdr_char'
out/target/product/dream/obj/lib/libgps.so: undefined reference to
`clnt_call'
out/target/product/dream/obj/lib/libgps.so: undefined reference to
`xdr_enum'
out/target/product/dream/obj/lib/libgps.so: undefined reference to
`xdr_bool'
out/target/product/dream/obj/lib/libgps.so: undefined reference to
`xdr_free'
out/target/product/dream/obj/lib/libgps.so: undefined reference to
`svc_register'
out/target/product/dream/obj/lib/libgps.so: undefined reference to
`xdr_long'
out/target/product/dream/obj/lib/libgps.so: undefined reference to
`svcerr_systemerr'
out/target/product/dream/obj/lib/libgps.so: undefined reference to
`xdr_short'
out/target/product/dream/obj/lib/libgps.so: undefined reference to
`xdr_u_char'
out/target/product/dream/obj/lib/libgps.so: undefined reference to
`xdr_u_long'
out/target/product/dream/obj/lib/libgps.so: undefined reference to
`xdr_opaque'
out/target/product/dream/obj/lib/libgps.so: undefined reference to
`xdr_string'
out/target/product/dream/obj/lib/libgps.so: undefined reference to
`xdr_u_int'
out/target/product/dream/obj/lib/libgps.so: undefined reference to
`xdr_u_quad_t'
out/target/product/dream/obj/lib/libgps.so: undefined reference to
`clnt_destroy'
out/target/product/dream/obj/lib/libgps.so: undefined reference to
`xdr_pointer'
out/target/product/dream/obj/lib/libgps.so: undefined reference to
`xdr_int'
out/target/product/dream/obj/lib/libgps.so: undefined reference to
`xdr_u_short'
collect2: ld returned 1 exit status
make: *** [out/target/product/dream/obj/EXECUTABLES/
AudioHardwareRecordLoop_intermediates/LINKED/AudioHardwareRecordLoop]
Error 1

Thanks.

Kindest regards,

[Disconnect]

Copy librpc.so to the same places libgps gets put (out/target/dream/out/lib iirc) and add it to the extract-files and android.mk

I'll upload a change to gerrit soon.

[Mike Lockwood]

Strange - that should not be necessary. librpc.so should be built
from the code in hardware/msm7k/librpc/

Mike

--

[Giam Teck Choon]

On Feb 2, 11:42 pm, Disconnect <dc.disconn...@gmail.com> wrote:
> Copy librpc.so to the same places libgps gets put (out/target/dream/out/lib
> iirc) and add it to the extract-files and android.mk
>
> I'll upload a change to gerrit soon.

Even I do that also ended up with the same error in case you wonder as
Mike stated already. I already did that before submit this error to
this list.

[sc430fc.localdomain:/home/choon/droid]$ find ./ -name librpc*
./out/target/product/dream/obj/include/librpc
./hardware/msm7k/librpc
[sc430fc.localdomain:/home/choon/droid]$ ls -al ./out/target/product/
dream/obj/include/librpc
total 12
drwxrwxr-x 3 choon choon 4096 2009-02-02 22:56 .
drwxrwxr-x 8 choon choon 4096 2009-02-02 22:56 ..
drwxrwxr-x 2 choon choon 4096 2009-02-02 22:56 rpc
[sc430fc.localdomain:/home/choon/droid]$ ls -al ./out/target/product/
dream/obj/include/librpc/rpc/
total 60
drwxrwxr-x 2 choon choon 4096 2009-02-02 22:56 .
drwxrwxr-x 3 choon choon 4096 2009-02-02 22:56 ..
-rw-rw-r-- 1 choon choon 6078 2009-02-02 22:56 clnt.h
-rw-rw-r-- 1 choon choon 1615 2009-02-02 22:56 pmap_clnt.h
-rw-rw-r-- 1 choon choon 1629 2009-02-02 22:56 rpc.h
-rw-rw-r-- 1 choon choon 1206 2009-02-02 22:56 rpc_router_ioctl.h
-rw-rw-r-- 1 choon choon 7364 2009-02-02 22:56 svc.h
-rw-rw-r-- 1 choon choon 13639 2009-02-02 22:56 types.h
-rw-rw-r-- 1 choon choon 5269 2009-02-02 22:56 xdr.h
[sc430fc.localdomain:/home/choon/droid]$ ls -al ./out/target/product/
dream/obj/include/librpc/
total 12
drwxrwxr-x 3 choon choon 4096 2009-02-02 22:56 .
drwxrwxr-x 8 choon choon 4096 2009-02-02 22:56 ..
drwxrwxr-x 2 choon choon 4096 2009-02-02 22:56 rpc

Thanks.

Kindest regards,

[Jean-Baptiste Queru]

I did see the same librpc.so build error this morning, but it seems
like a clean build (rm -rf out) fixed it.

To be double-sure, I downloaded a clean source tree, followed the
dream build instructions as per the web site, did a build, and it went
fine.

I guess it's just some transient glitch that disappears when doing a
clean build.

As far as I'm concerned, case closed.

JBQ

--

[Giam Teck Choon]

On Feb 2, 11:54 pm, Mike Lockwood <lockw...@android.com> wrote:
> Strange - that should not be necessary. librpc.so should be built
> from the code in hardware/msm7k/librpc/

Ok Mike I successfully completed the build/compile ;)
To summarize it is actually:

$ make clean && make dataclean && make installclean
$ make modules librpc
$ make

I guess sub-modules librpc is not build at all from default make build
option (see my previous message above about find librpc output). Can
the in-charge take a look into this when available?

Lengthy one below... ...

I successfully complete the build by using extra steps without
patching so I guess it is modules librpc not being build fully
somewhere... ... Sorry for the below lengthy message but I guess this
might help someone.

This is how I complete the build/compile:

Create a new folder and repo sync the source with htc/dream etc. Then
run:

$ make clean && make dataclean && make installclean

Then run:
$ make showcommands

And you will get those errors I posted previously... ...

SNAP...

collect2: ld returned 1 exit status
make: *** [out/target/product/dream/obj/EXECUTABLES/
AudioHardwareRecordLoop_intermediates/LINKED/AudioHardwareRecordLoop]
Error 1

Now here is how I continue by reading the Makefile which includes
build/core/main.mk. Since librpc is not being compile or build fully
so after looking build/core/main.mk I noticed there is an option to
build modules/Available sub-modules so I run:

$ make modules

And I get the list of available sub-modules shown below:

[sc430fc.localdomain:/home/choon/droid]$ make modules
build/core/product_config.mk:211: WARNING: adding test OTA key
============================================
TARGET_PRODUCT=htc_dream
TARGET_BUILD_VARIANT=eng
TARGET_SIMULATOR=
TARGET_BUILD_TYPE=release
TARGET_ARCH=arm
HOST_ARCH=x86
HOST_OS=linux
HOST_BUILD_TYPE=release
BUILD_ID=
============================================
build/core/main.mk:180: implicitly installing apns-conf_sdk.xml
build/core/copy_headers.mk:15: warning: overriding commands for target
`out/target/product/dream/obj/include/libpv/getactualaacconfig.h'
build/core/copy_headers.mk:15: warning: ignoring old commands for
target `out/target/product/dream/obj/include/libpv/
getactualaacconfig.h'
build/core/Makefile:17: warning: overriding commands for target `out/
target/product/dream/system/etc/apns-conf.xml'
build/core/Makefile:17: warning: ignoring old commands for target `out/
target/product/dream/system/etc/apns-conf.xml'
build/core/Makefile:17: warning: overriding commands for target `out/
target/product/dream/system/etc/wifi/tiwlan.ini'
build/core/base_rules.mk:384: warning: ignoring old commands for
target `out/target/product/dream/system/etc/wifi/tiwlan.ini'
Available sub-modules:
01-test libbluez-utils-common-static libsoundpool
20-dns.conf libbz libsqlite
95-configured libc libsqlite3_android
aapt libcamera libSR_AcousticModels
acp libcameraservice libSR_AcousticState
adb libc_common libSR_AudioIn
adbd libc_debug libSR_Core
add-property-tag libclearsilver-jni libsrec_jni
afar libcolorconvert libSR_EventLog
aidl libcorecg libSR_G2P
AlarmClock libcpm libSR_Grammar
AliasActivity libcrypto libSR_Nametag
am libctest libSR_Recognizer
amend libCustomWifi libSR_Semproc
android libcutils libSR_Session
android.awt libdb libSR_Vocabulary
android.policy libdbus libssl
android.policy_mid libdbus-tools-common libstdc++
android.policy_phone libdex libsurfaceflinger
androidprefs libdl libswt-atk-gtk-3236
android.test.runner libdrm1 libswt-cairo-gtk-3236
AndroidTests libdrm1_jni libswt-gtk-3236
angeles libdrm2 libswt-pi-gtk-3236
ant libdvm libsystem_server
anttasks libebl libthread_db
apicheck libebl_arm libthreadsafe_callback_ao
ApiDemos libeglib_static libtinyxml
ApiDemosTests libelf libtomcrypt
apkbuilder libelfcopy libtommath
applypatch libembunit libui
app_process libESR_Portable libunit_test
apriori libESR_Shared libunz
asm-3.1 libexif libutil
atree libexpat libutils
attest libext libvorbisidec
audio libfdlibm libwbxml
AudioHardwareRecord libFFTEm libwbxml_jni
AudioHardwareRecordLoop libfst libwebcore
AudioInRecord libft2 libWifiApi
auth-agent libgdbus_static libwpa_client
avinfo libgetactualaacconfig libxml2
bb2sym libgif libxml2wbxml
bb_dump libGLES_CM libz
bbprof libgoogleclient libzipfile
bison libhardware line_endings
bookmarks.xml libhardware_legacy linker
brfpatch libhcid localize
Browser libhost location
browsertestplugin libicudata logcat
bsdiff libicui18n logwrapper
bspatch libicuuc lsd
bttest libiptc LunarLander
Calculator libjavacore LunarLanderTests
CalculatorTests libjpeg make_cfst
Calendar liblog make_g2g
CalendarProvider libm makekeycodes
CalendarProviderTests libm4v_config make-update-script
CalendarTests libmedia make_ve_grammar
Camera libmedia_jni mediaframeworktest
CameraTests libmediaplayerservice MediaProvider
check-lost+found libmincrypt mediaserver
check_prereq libminui memtest
check_trace libminzip mkbootfs
clearsilver libmp4recognizer_utility mkbootimg
cmu2nuance libmtdutils mksdcard
com.android.im.plugin libnativehelper mkyaffs2image
com.example.android.platform_library libneo_cgi Mms
com.example.android.platform_library.xml libneo_cs monkey
Compass libneo_util mountd
Contacts libnetutils Music
ContactsProvider libomx_aac_component_lib MusicTests
ContactsTests libomx_aacdec_sharedlibrary nc
core libomx_amr_component_lib netcfg
core-tests libomx_amrdec_sharedlibrary netperf
CoreTests libomx_amrenc_component_lib netserver
coverage libomx_amrenc_sharedlibrary ninepatch
cpueater libomx_avc_component_lib nmea
crasher libomx_avcdec_sharedlibrary NotePad
create_test_dmtrace libomx_avcenc_component_lib NotePadTests
daemonize libomx_avcenc_sharedlibrary opcontrol
dalvikvm libomx_baseclass_lib openssl
dbus-daemon libomx_common_lib oprofiled
dbus-monitor libomx_m4v_component_lib
org.eclipse.core.commands_3.2.0.I20060605-1400
dbus-send libomx_m4vdec_sharedlibrary
org.eclipse.equinox.common_3.2.0.v20060603
ddmlib libomx_m4venc_component_lib
org.eclipse.jface_3.2.0.I20060605-1400
ddms libomx_m4venc_sharedlibrary org-netbeans-api-visual
ddmuilib libomx_mastercore_lib org-openide-util
debuggerd libomx_mp3_component_lib PackageInstaller
Development libomx_mp3dec_sharedlibrary pand
dexdump libomx_queue_lib parseStringTest
dexlist libomx_sharedlibrary passkey-agent
dexopt libop perm_checker
dexopt-wrapper libopencore_2way perm_checker.conf
dexpreopt libopencore_author Phone
dhcpcd libopencore_common ping
dhcpcd.conf libopencore_download PlatformLibraryClient
dhcpcd-run-hooks libopencore_downloadreg platform.xml
dictTest libopencore_mp4local pm
DisabledTestApp libopencore_mp4localreg post_trace
dmtracedump libopencore_net_support pppd
dosfsck libopencore_player preload
DownloadProvider libopencore_rtsp procmem
draw9patch libopencore_rtspreg procrank
DrmProvider libosclbase profile_pid
droiddoc libosclerror profile_trace
dumpeventlog libosclio properties
dumpkey liboscllib pv2way_omx_engine_test
DumpRenderTree libosclmemory pvplayer
dumpstate libosclproc pvplayer_engine_test
dumpsys libosclregcli q2dm
dund libosclregserv q2g
dvz libosclutil qemud
dx libpagemap qwerty2.kcm
Email libpassthru_oma1 qwerty.kcm
EmailTests libpcap radiooptions
emma libpixelflinger rctest
emulator libpixelflinger_armv6 read_addr
emulator-arm libpixelflinger_static read_method
emulator-dyngen libplatform_library_jni read_pid
emulator-op libpng read_trace
emulator-op-trace libpopt recovery
EnabledTestApp libprotocolenginenode rfcomm
eventanalyzer libpv2wayengine rgb2565
event-log-tags libpv324m rild
exc_dump libpv_aac_dec RSSReader
ext libpvaacffparsernode run-core-tests
Fallback libpvaacffrecognizer safe_iop_test
fastboot libpvaacparser scotest
favorites.xml libpvamrffparsernode scp
flash_image libpvamrffrecognizer screenshot2
FontLab libpv_amr_nb_common_lib sdklib
framework libpvamrwbdecoder sdkmanager
framework-res libpvauthorengine SdkSetup
FrameworkTest libpv_avc_common_lib sdkstats
framework-tests libpvavcdecoder sdkuilib
FrameworkTestTests libpvavch264enc sdptool
fs_config libpvavifileparser sdutil
fs_get_stats libpvcommsionode service
GadgetHost libpv_config_parser servicemanager
gdbserver libpvdecoder_gsmamr services
genext2fs libpvdownloadinterface Settings
GlobalTime libpvdownloadmanagernode SettingsProvider
GoogleContactsProvider libpvdownloadreginterface SettingsTests
GoogleContactsProviderTests libpvencoder_gsmamr sh
google-framework libpvfileoutputnode showlease
googlelogin-client libpvfileparserutils showmap
GoogleSearch libpvframemetadatautility showslab
GoogleSubscribedFeedsProvider libpvgendatastruct signapk
gpustate libpvgeneraltools SimpleJNI
grxmlcompile libpvgsmamrparser SkeletonApp
gzip libpv_http_parcom SkeletonAppTests
hciattach libpvid3parcom SmokeTest
hciconfig libpvjitterbuffernode SmokeTestApp
hcid libpvlatmpayloadparser Snake
hcidump libpvm4vencoder SnakeTests
hcitool libpvmediadatastruct SoftKeyboard
HelloActivity libpvmediainputnode soslim
HelloActivityTests libpvmedialayernode sound
hierarchyviewer libpvmediaoutputnode SoundRecorder
hist_trace libpvmf SpareParts
Home libpvmfrecognizer sqlite3
hstest libpvmimeutils SRecTest
HTMLViewer libpvmioaviwavfileinput SRecTestAudio
icache libpvmio_comm_loopback ssh
icudata libpvmiofileinput SslLoad
idegen libpvmiofileoutput stack_dump
IM libpvmp3 StatusBarTest
ime libpvmp3ff Stk
ImfTest libpvmp3ffparsernode strace
ImProvider libpvmp3ffrecognizer su
init libpvmp4decoder SubscribedFeedsProvider
input libpvmp4ff surfaceflinger
installd libpvmp4ffcomposer svc
iptables libpvmp4ffcomposernode swing-worker-1.1
iself libpvmp4ffparsernode swt
isprelinked libpvmp4ffrecognizer Sync
jarjar libpvmp4interface system_server
jarutils libpvmp4reginterface tcpdump
jasmin libpvomxaudiodecnode TelephonyProvider
jasmin.jar libpvomxbasedecnode temp_layoutlib
jcommon-1.0.12 libpvomxencnode Term
jdiff libpv_omx_interface test-fb-refresh
jdwpspy libpvomx_proxy_lib test-fb-simple
jfreechart-1.0.9 libpvomxvideodecnode test_g2g
jfreechart-1.0.9-swt libpvplayer_engine test-opengl-codegen
kcm libpvrtsp_cli_eng_node test-opengl-filter
kxml2-2.3.0 libpvrtspinterface test-opengl-finish
l2ping libpv_rtsp_parcom test-opengl-sfsim
l2test libpvrtspreginterface test-opengl-textures
latencytop libpvsdpparser test-opengl-tritex
Launcher libpvsocketnode test_pvauthorengine
launchperf libpvstreamingmanagernode test_swiarb
layoutlib libpvthreadmessaging test_zipfile
layoutlib_api libpvvideoparsernode timeinfo
layoutlib_create libpvwav tiwlan.ini
layoutlib_utils libpvwavffparsernode toolbox
liba2dp libpvwavffrecognizer traceview
libabi librank trout-keypad-qwertz.kcm
libaes libreference-ril trout-keypad-v3.kcm
libagl libril Updater
libamend librpc usbtest
libandroidpv librtppayloadparser VoiceDialer
libandroidpvauthor librtprtcp wbxmltest
libandroid_runtime libsafe_iop wdsclient
libandroid_servers libSDL wlan_cu
libarity libSDLmain wlan_loader
libaudio libsdpserver_static wpa_cli
libaudioflinger libsgl wpa_supplicant
libaudiointerface libsimplejni wpa_supplicant.conf
libbluedroid libskiagl zipalign
libbluetooth libsonivox
======================================================================

Now if I run:
$ make modules AudioHardwareRecord
Will get the same error:

make: *** [out/target/product/dream/obj/EXECUTABLES/
AudioHardwareRecordLoop_intermediates/LINKED/AudioHardwareRecordLoop]
Error 1

Now I run:
$ make modules librpc
Will show me the sub-modules listing message then follow by:

target thumb C: librpc <= hardware/msm7k/librpc/xdr.c
target thumb C: librpc <= hardware/msm7k/librpc/rpc.c
target thumb C: librpc <= hardware/msm7k/librpc/svc.c
target thumb C: librpc <= hardware/msm7k/librpc/clnt.c
hardware/msm7k/librpc/clnt.c: In function 'clnt_create':
hardware/msm7k/librpc/clnt.c:568: warning: passing argument 1 of
'memset' discards qualifiers from pointer target type
target thumb C: librpc <= hardware/msm7k/librpc/ops.c
target thumb C: librpc <= hardware/msm7k/librpc/svc_clnt_common.c
target StaticLib: librpc (out/target/product/dream/obj/
STATIC_LIBRARIES/librpc_intermediates/librpc.a)
target SharedLib: librpc (out/target/product/dream/obj/
SHARED_LIBRARIES/librpc_intermediates/LINKED/librpc.so)
target Prelink: librpc (out/target/product/dream/symbols/system/lib/
librpc.so)
target Strip: librpc (out/target/product/dream/obj/lib/librpc.so)
Notice file: hardware/msm7k/librpc/NOTICE -- out/target/product/dream/
obj/NOTICE_FILES/src//system/lib/librpc.so.txt
Notice file: bionic/libc/NOTICE -- out/target/product/dream/obj/
NOTICE_FILES/src//system/lib/libc.so.txt
Notice file: bionic/libdl/NOTICE -- out/target/product/dream/obj/
NOTICE_FILES/src//system/lib/libdl.so.txt
Install: out/target/product/dream/system/lib/libdl.so
Notice file: bionic/libc/NOTICE -- out/target/product/dream/obj/
NOTICE_FILES/src//system/lib/libc_common.a.txt
Install: out/target/product/dream/system/lib/libc.so
Notice file: bionic/libstdc++/NOTICE -- out/target/product/dream/obj/
NOTICE_FILES/src//system/lib/libstdc++.so.txt
Install: out/target/product/dream/system/lib/libstdc++.so
Notice file: bionic/libm/NOTICE -- out/target/product/dream/obj/
NOTICE_FILES/src//system/lib/libm.so.txt
Install: out/target/product/dream/system/lib/libm.so
Install: out/target/product/dream/system/lib/liblog.so
Notice file: hardware/msm7k/librpc/NOTICE -- out/target/product/dream/
obj/NOTICE_FILES/src//system/lib/librpc.a.txt
Install: out/target/product/dream/system/lib/librpc.so

Finally run below command will be successful:
$ make modules AudioHardwareRecordLoop

Then to finish/complete you can actually skip the make modules
AudioHardwareRecordLoop as long as librpc.so is being built/install to
target and just do a make instead :)

Hope the above information is useful to someone ;)

Thanks.

Kindest regards,

[Giam Teck Choon]

On Feb 3, 6:42 am, Jean-Baptiste Queru <j...@google.com> wrote:
> I did see the same librpc.so build error this morning, but it seems
> like a clean build (rm -rf out) fixed it.

That didn't fix for me during that time before the build :(

> To be double-sure, I downloaded a clean source tree, followed the
> dream build instructions as per the web site, did a build, and it went
> fine.

I guess I shall learn to wait before posting such in future maybe a
few days later as past two days at least I unable to build
successfully with each try doing a clean source tree first... ...

And now I got the following:

[sc430fc.localdomain:/home/choon/droid]$ repo sync
remote: Counting objects: 7, done.
remote: Compressing objects: 100% (4/4), done.
remote: Total 4 (delta 3), reused 0 (delta 0)
Unpacking objects: 100% (4/4), done.
From git://android.git.kernel.org/platform/build
aeb2fa6..4cf7606 master -> korg/master

> I guess it's just some transient glitch that disappears when doing a
> clean build.

Mostly... ...

> As far as I'm concerned, case closed.
>
> JBQ
>

Thanks for your time to verify such build issue. Greatly
appreciate ;)

Kindest regards,

[Jean-Baptiste Queru]

As far as I can tell HW GL seems to be working (or at least globaltime
suggests so).

JBQ

--

[oth...@googlemail.com]

Jean-Baptiste Queru wrote:
> As far as I can tell HW GL seems to be working (or at least globaltime
> suggests so).

Works for me. I now got HW GL and Bt for the first time from the OS
repository.
Great!
I had one little build issue (see below) after doing a super clean
checkout. But copying librpc.so to out/target/product/dream/obj/lib/
from a previous build folder fixed this for me.
Tim

target Executable: AudioHardwareRecordLoop (out/target/product/dream/
obj/EXECUTABLES/AudioHardwareRecordLoop_intermediates/LINKED/
AudioHardwareRecordLoop)

/Volumes/hfsAndroid/cupcake/prebuilt/darwin-x86/toolchain/arm-
eabi-4.2.1/bin/../lib/gcc/arm-eabi/4.2.1/../../../../arm-eabi/bin/ld:
warning: librpc.so, needed by out/target/product/dream/obj/lib/

[Disconnect]

I think the changes went in last night to fix that, but if not "make librpc ; make {whatever}" works also, and uses the newest version of librpc.

[Jean-Baptiste Queru]

Change 8761 is still up for review (but I expect it to go in sometime
today). In a nutshell, there's a missing dependency in the build
system, such that librpc isn't currently always built as early as it
should.

Like Disconnect said, building librpc explicitly first is the best
workaround until the fix goes in.

JBQ

--

[oth...@googlemail.com]

This is not fully clear to me yet. Are all code changes, these days,
going through Gerrit? Or is some code (maybe completely new code)
getting in to the repository through some other channels?
Tim

[Jean-Baptiste Queru]

The drops from google's internal servers have been done so far through
a backdoor and do not show up on Gerrit.

JBQ

--

[Pascal Merle]

I doubt the build fully includes hw graphics acceleration. Although
there are no obvious delays in the global time sample overall graphics
performance is poor. E.g. window animations take long, need to be
turned off. Blurring of backgrounds for alert messages are so slow
that you can watch each single darkening step. This gets confirmed by
logcat during startup:

02-12 22:52:12.601: INFO/SurfaceFlinger(64): SurfaceFlinger's main
thread ready to run. Initializing graphics H/W...
02-12 22:52:12.761: DEBUG/SurfaceFlinger(64): pid 64 requesting gpu
core (owner = -1)
02-12 22:52:12.761: WARN/SurfaceFlinger(64): couldn't grant gpu core
to pid 64
02-12 22:52:12.761: DEBUG/GLLogger(64): requestGPU returned -1
02-12 22:52:12.761: ERROR/GLLogger(64): h/w accelerated eglGetDisplay
() failed (EGL_SUCCESS)

To get rid of the problem I tried build variant "userdebug" (instead
of "eng"), same messages as above but does not boot beyond the
blinking android :-(
Also I noticed that there is a problem with DHCP currently. Everytime
wifi is reconnected after being dormant another dhcpd process gets
added. It takes about 10 seconds to resolve dhcp. First a file lock on
the dhcp pid file is logged, then on the second attempt dhcp runs but
the old dhcpd is still there.