Re: [boost] Boost regression testing on Android

Subject: Re: [boost] Boost regression testing on Android
From: Dmitry Moskalchuk (dm_at_[hidden])
Date: 2015-02-04 07:59:06

Well, test script itself is at
https://github.com/crystax/android-platform-ndk/blob/master/tests/run-boost-tests.
It is part of CrystaX NDK file tree and it do many things to setup
environment for Boost testing. As you can see, it also generate several
shell script wrappers to be able run tests; one of such wrappers is
adbrunner. Content of regression/armeabi-v7a-hard folder is fully
generated on the fly by run-boost-tests script.

However, I'd recommend to not use this script right now since I'm going
to refactor it a bit - at least, extract adbrunner and do several other
changes. Please wait several days and I'll provide full sources and
instructions on how to run Boost regression testing on Android.

BTW, if you'll look on
http://boost.crystax.net/master/developer/summary.html, you'll see clang
toolchains indicated as 'clang-linux-3.4' and 'clang-linux-3.5'. It goes
from the fact that Boost build scripts suppose there are just two types
of clang - clang-darwin and clang-linux, and switch between them based
on host OS condition, which is wrong. It should be based on target OS.
Look here:
https://github.com/boostorg/build/blob/develop/src/tools/clang.jam. I
suppose there should be smth like this:

if $(target-os) = darwin
    toolset.using clang-darwin .....
else if $(target-os) = linux
    toolset.using clang-linux ...
else if $(target-os) = android
    toolset.using clang-android ....

Or smth like that. I don't know bjam syntax good, so it's just to
indicate my idea. I'd be appreciate if you could fix this while I'm
finishing with scripts refactoring.

Finally, when I finish with my task, we could modify Boost jam scripts
to add support of NDK compilers directly instead of relying on wrappers
for g++/clang++.

-- 
Dmitry Moskalchuk
On 04/02/15 14:17, Antony Polukhin wrote:
> 2015-01-21 21:00 GMT+04:00 Dmitry Moskalchuk <dm_at_[hidden]>:
>
>> Sure, we have HTML results of regression tests produced by
>> library_status utility; I just need advice how to integrate those
>> results into testing matrix you've mentioned.
>> Maybe, some article about that?
>>
> I'm very interested in your adbrunner and run.py scripts. Whole
> regression/armeabi-v7a-hard folder looks interesting to me.
>
> Could you please provide the sources? I'll do my best to integrate your
> work into the Boost regression testing suit and will help you with docs.
>