[7zip] restore missing sources and fix compilation on android

[msclock]

Fixes #42910

References #41721, #41743

[msclock]

@DanAlbert in #41721, for This fixes compilation errors with NDK r28., Could you explain your compilation errors in detail. Thanks.

Still skip failures from Android or do more investigation🤔🤔

[dg0yt]

Why should the sources in the Windows directory be added to a non-WIN32 build?

[msclock]

Why should the sources in the Windows directory be added to a non-WIN32 build?

@dg0yt see #42910

[dg0yt]

@dg0yt see #42910

I have seen that already.

[msclock]

Why should the sources in the Windows directory be added to a non-WIN32 build?

AFAIK, 7zip uses win32 com api to implement a plugin loading way, so it implements the same api on linux. And Some api are placed in that Windows directory and refered by sources under the directory CPP/7zip/Archive with relative paths.

But in detail should ask the author of the initial port PR.

[msclock]

@DanAlbert in #41721, for This fixes compilation errors with NDK r28., Could you explain your compilation errors in detail. Thanks.

Still skip failures from Android or do more investigation🤔🤔

@dg0yt @DanAlbert The failure is timespec_get is broken on android and I have handled it.

[dg0yt]

The failure is timespec_get is broken on android and I have handled it.

This may be too simplified.

• The amount of API offered by Android varies with API level.
• Recent NDK updates try to improve how the availability of implementations is presented during compilation. However, this has compatibility trade-offs. In particular when packages are not maintained with Android in mind (or in CI).

[msclock]

Have added constratins __ANDROID_API__ refered from tinycthread/tinycthread#47

[WangWeiLin-MV]

The port installation tests pass with the following triplets:

• x64-android
• x64-linux

[DanAlbert]

iirc the problem was that those files call APIs that don't exist on Android (or at least didn't at the minSdkVersion the port used, but I think it was the former)