Split Initializer into 2 source files to allow 32-bit debug build

[metorm]

There is a trouble in src/Initializer.cc. Since the class includes a giant recursive template function Initializer::DefineExternFunction. Compiling 32-bit debug version produced so huge object file that as.exe throws "too many sections & file too big" error. I have to split the class into 2 files to evade the error.

See file src/initializer2.cc

[CLAassistant]

Thank you for your submission, we really appreciate it. Like many open source projects, we ask that you sign our Contributor License Agreement before we can accept your contribution.
_{You have signed the CLA already but the status is still pending? Let us recheck it.}

[rakhimov]

Thanks @metorm ! Indeed, the Initializer is complex, but the proposed quick-&-dirty splitting is not a proper solution since each module must be represented by a single header and implementation file.

This issue with debug info blow-up will show-up again when scram switches to Verdigris, which heavily uses a lot of template & compile time features. I would recommend trying with reduced debug info -g1 flag. Since 32bit archs are only target release platforms, I wouldn't worry much with debug builds.

[metorm]

OK, I understand that. I'll keep that in my fork since I may have to debug something on the 32Bit platform.

BTW, with my MSYS2, jemalloc throws segment fault in both 32/64Bit, debug and release version. I cannot figure out the reason.

[rakhimov]

BTW, with my MSYS2, jemalloc throws segment fault in both 32/64Bit, debug and release version. I cannot figure out the reason.

Yea, there is some strange issue mingw jemalloc: https://github.com/Alexpux/MINGW-packages/issues/2520#issuecomment-390658023

I'd recommend disabling it for now with -DWITH_JEMALLOC=OFF.

[metorm]

I already did that.

In addition, I have translated the program to Simplified Chinese at my fork. However, my fork and the original version is now out of sync, so I cannot create a proper PR. You can fetch the translation using cherry-pick if you like.

[rakhimov]

@metorm Please submit translations through Crowdin.

[rakhimov]

Also, you should not use the upstream branches in your development. As stated in the contributing guidelines, you must create topic branches and PR with them.

[metorm]

En…

[rakhimov]

En…

? @metorm

[metorm]

I mean, all right. Contributing to an open source project seems to require more experience than I thought.