Some improvements for packaged build [wip]

[noiseless]

I would like to share you my suggestions for improving TG_OWT_PACKAGED_BUILD.

Here they are:

1. cherry-picked include cmake rules for third party libraries only if it needed #66 && fixed merge conflict
2. This patch is an adaptation of Vitaly Lipatov's work [email protected] for the latest versions of tg_owt. These changes are a credit to him and are now, as far as I can see, already being used to build alt linux packages.

The patch adds cmake rules for building tg_owt with libyuv from the system/packages.

3. In this patch I fix the paths to the libyuv header files. Similar patch for tgcalls (separately they make little sense - packaged build will break).

Testing.

To test my changes I did the following:

1. I have done the build according to the "Build instructions for Linux using Docker".

To apply my changes, I replaced these lines in centos_env/Dockerfile from

RUN git remote add origin $GIT/desktop-app/tg_owt.git
RUN git fetch --depth=1 origin b02478677baac6d563589f216800ff9cea0fd65c

to

RUN git remote add origin $GIT/noiseless/tg_owt.git
RUN git fetch --depth=1 origin c5fc0b3ac814b599c161fcc93e62e9a267c86662

and also edited .gitmodules:

[submodule "Telegram/ThirdParty/tgcalls"]
    path = Telegram/ThirdParty/tgcalls
    url = https://github.com/noiseless/tgcalls.git
    branch = system-libyuv

The build was successful.
2) I did a packaged build under OpenBSD. Tested situations where libyuv is installed on the system and where it is not there, but everything you need is in ./src/third_party/libyuv.

In the first case, tg_owt was built with libyuv from the package, in the second, from ./src/third_party/libyuv, i.e. everything is as expected.

What do you think of my patches? If necessary, I am ready to improve them according to your wishes.
Perhaps I should test some other build variants?

Thanks anyway.

[ilya-fedin]

• In the last patch I turn off cmake/{libabsl,libopenh264,libusrsctp,libvpx,libyuv}.cmake for packaged builds.

Oh. Every time someone tries to do this, snap package breaks. So this would need extensive testing across all the building configuration.

[ilya-fedin]

This breaks the fallback logic when a library is not found

[klemensn]

Yes, that can't be right.
#85 should do the trick, though.

[noiseless]

Yes, this is wrong, as Ilya has already told me. 5986975 should fix that.
(However, there are linking problems with snap pack builds there; I haven't had a chance to reproduce it yet, but I'm still going to)

[ilya-fedin]

Is there still any work? Maybe close the PR and re-open when you would have time?

[noiseless]

reasonably