[Ohrrpgce] android build machine

James Paige Bob at hamsterrepublic.com
Mon Dec 25 14:31:24 PST 2023 

I have no idea how I managed to get platform version 31 installed in my
2012 Android SDK *shrug*
I also ran into a "Source option 5 is no longer supported." openjdk8 was
the last java to support whatever "option 5" means. And Debian 9 was the
last debian to support openjdk8

After a lot of messing around, I think I finally got the android nightly
building in a docker image (still using my cargo-cult artifact copy of the
android SDK)

I'll push the Dockerfile for it. I ended up using Debian 11, but installing
an Amazon "corretto" backport of OpenJDK-8 and I /think/ it all works. I
managed to produce a debug apk a few minutes ago. If I can confirm it
actually installs and runs on my phone I'm going to update the nightly
build system to do it this way.

> And I notice that block was modified in 4701548e0 ("SDL: switched build
system from Ant to Gradle") to make "android" optional. It's also used in
build/envsetup.sh but I don't remember ever running that.

Aha, thanks! I'll check that out too.



On Mon, Dec 25, 2023 at 4:56 PM Ralph Versteegen <teeemcee at gmail.com> wrote:

> Well, you don't want the SDK that I have, because it no longer works!
> Running "android" (a binary from 2018) I get Android SDK Manager revision
> 25.2.5 which shows the most recent "SDK Platform" available for install
> being one for Android API 29 (Android 10). Which is the one I have
> installed. I had been meaning to install a newer SDK but had no idea that
> the SDK manager itself was no longer functional.
>
> Since you have targetSdkVersion set to 30, my SDK no longer works (it gets
> all the way through compiling and linking but fails in the final steps of
> packaging the .apk). I managed to create an .apk only by decreasing the
> target to 29 and changing my PATH to point to an older javac (javac from
> jdk-17.0.1 threw an error "Source option 5 is no longer supported. Use 7 or
> later." (about its "-source 1.5" argument), while javac 1.8.0 from
> openjdk-8u392 only printed a deprecation warning. Seems like Java has
> suffered a Perl 5/6-style fork)
>
> BTW when I try to edit  android:targetSdkVersion="30" in
> project/AndroidManifest[Template].xml it seems to have no effect. I finally
> figured out I have to edit project/project.properties too. That file claims
> to be autogenerated but is it? I couldn't figure out how to do so.
>
>
>
> On Tue, 26 Dec 2023 at 06:11, James Paige via Ohrrpgce <
> ohrrpgce at lists.motherhamster.org> wrote:
>
>> TMC, Question! What version of the Android SDK do you have installed?
>>
>> I think my version is probably from around 2012, and it is so old I don't
>> even know how to install it again (which makes it rough to create a Docker
>> image for it)
>>
>> I tried with the latest SDK, but it doesn't even have the "android"
>> binary anymore, it has "sdkmanager" instead with different command line
>> arguments, but our dusty old fork of sdl-android relies on "android"
>>
>> I'm hoping to find the oldest sdk that still works as a quick-fix, but
>> googling for "when did android SDK stop having android" doesn't work very
>> well :D
>>
>>
>>
>> On Thu, Dec 21, 2023 at 9:10 PM James Paige <Bob at hamsterrepublic.com>
>> wrote:
>>
>>> Oh, I am pretty sure I know what is wrong. The nightly build vm for
>>> android is still Debian 9. It has scons version 2.3, which is still python
>>> 2.7 based. The latest scons on my Debian 11 box is 4.0
>>>
>>>
>>>
>>> On Tue, Dec 19, 2023 at 9:13 PM James Paige <Bob at hamsterrepublic.com>
>>> wrote:
>>>
>>>>
>>>>
>>>> On Sat, Dec 16, 2023 at 7:40 PM Ralph Versteegen via Ohrrpgce <
>>>> ohrrpgce at lists.motherhamster.org> wrote:
>>>>
>>>>> That's strange... works on my machine. The error is:
>>>>>
>>>>> jni/../jni/application/ohrrpgce/tmp//filelayer.cpp:7:24: fatal error:
>>>>> fb/fb_stub.h: No such file or directory
>>>>>  #include "fb/fb_stub.h"
>>>>>                         ^
>>>>>
>>>>> And the reason for the error is that copy_source_actions() in
>>>>> ohrbuild.py isn't copying fb/fb_stub.h to android/tmp/ apparently because
>>>>> node.get_implicit_deps(), which is meant to run a sconscript builtin source
>>>>> scanner to return the list of headers isn't working. Try uncommenting the
>>>>> two lines immediately below that:
>>>>>         def scstr(x): return ",".join(str(y) for y in x)
>>>>>         print("node", str(node), "sources", scstr(node.sources),
>>>>> "headers", scstr(headers))
>>>>> It should print:
>>>>> node filelayer.cpp sources  headers
>>>>> errorlog.h,fb/fb_stub.h,filelayer.hpp,lumpfile.h,mutex.hpp,config.h,miscc.h,fb/fb_array.h,fb/fb_config.h,fb/fb_device.h,fb/fb_error.h,fb/fb_file.h,fb/fb_string.h,fb/fb_thread.h,os.h,fb/fb_blackbox_config.h
>>>>>
>>>>
>>>> What it actually prints on the nightly build machine is:
>>>>
>>>> node filelayer.cpp sources  headers
>>>>
>>>> I haven't spotted why yet.
>>>>
>>>>
>>>>> As for .aab support... I had a look at upstream commandergenius. They
>>>>> seem to use gradle to create the .aab file, plus have a script for signing
>>>>> it. Gradle isn't used for just that. The addition of all the gradle stuff
>>>>> is new since we forked. I don't know whether you want to copy or reuse the
>>>>> upstream work, requiring gradle, or completely reimplement support some
>>>>> other way.
>>>>>
>>>>> Trying to merge our fork with upstream commandergenius with "git
>>>>> merge" isn't going to work ("Your branch and 'origin/sdl_android' have
>>>>> diverged, and have 79 and 1915 different commits each, respectively" which
>>>>> doesn't begin to describe the enormous patches and conflicts... I tried git
>>>>> merge after which on the first conflict "git status" outputs 17972 lines
>>>>> and "git diff" 10588 lines of conflicts!). Manually merging it by
>>>>> cherry-picking or applying (some of) our commits on top of upstream might
>>>>> be practical but is still tons of work. Cherry-picking their commits which
>>>>> add .aab support is clearly far less work but isn't easy either, because on
>>>>> top of the .aab signing commits (I couldn't easily figure out which commits
>>>>> are needed but "git log --grep bundle" is a good start) you'd also need all
>>>>> the ones for gradle support.
>>>>>
>>>>
>>>> First I'll dig around and see if there is a way to just change the
>>>> existing scripts to support aab format, which might be possible, I don't
>>>> know yet.
>>>> If not, I think trying to cherry-pick our commits on top of upstream
>>>> sounds like the least work, but I'll cross that bridge only if I don't find
>>>> any easier shortcuts.
>>>>
>>>> I'll keep you posted if I find anything.
>>>>
>>>>
>>>>>
>>>>> On Sat, 16 Dec 2023 at 01:49, James Paige via Ohrrpgce <
>>>>> ohrrpgce at lists.motherhamster.org> wrote:
>>>>>
>>>>>> I forgot to mention, the android builds have been broken for a while
>>>>>> (Since October 27)
>>>>>>
>>>>>> I haven't had time to narrow it down to the exact revision yet, but
>>>>>> I'll work on it when I have time.
>>>>>> I also want to work on the Android 12 fix, (I know what to do) and
>>>>>> the .aab format (Don't know what to do yet, but I'll figure it out
>>>>>> eventually)
>>>>>>
>>>>>> ---
>>>>>> James
>>>>>>
>>>>>> _______________________________________________
>>>>>> Ohrrpgce mailing list
>>>>>> ohrrpgce at lists.motherhamster.org
>>>>>> http://lists.motherhamster.org/listinfo.cgi/ohrrpgce-motherhamster.org
>>>>>>
>>>>> _______________________________________________
>>>>> Ohrrpgce mailing list
>>>>> ohrrpgce at lists.motherhamster.org
>>>>> http://lists.motherhamster.org/listinfo.cgi/ohrrpgce-motherhamster.org
>>>>>
>>>> _______________________________________________
>> Ohrrpgce mailing list
>> ohrrpgce at lists.motherhamster.org
>> http://lists.motherhamster.org/listinfo.cgi/ohrrpgce-motherhamster.org
>>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.motherhamster.org/pipermail/ohrrpgce-motherhamster.org/attachments/20231225/a68acd4e/attachment.htm>

More information about the Ohrrpgce mailing list