Prep for upstream llvm change to main mangling. NFC
#17099
Conversation
This change should allow https://reviews.llvm.org/D75277 to land without breaking emscripten. In the long run I hope to remove all of this in favor of calling `_start` from JS, but that can happen later, and could be user-visible if folks are calling `main` themselves.
| @@ -1741,6 +1741,9 @@ def phase_linker_setup(options, state, newargs, user_settings): | |||
| # Note the exports the user requested | |||
| building.user_requested_exports.update(settings.EXPORTED_FUNCTIONS) | |||
|
|
|||
| if '_main' in settings.EXPORTED_FUNCTIONS: | |||
| settings.EXPORT_IF_DEFINED.append('__main_argc_argv') | |||
There was a problem hiding this comment.
A comment here would be good. This stuff is adding a bunch of complexity and comments can at least help the reader with that. The other locations at least use the MANGLED_MAIN setting that the reader can then look up, but this location doesn't even have that.
There was a problem hiding this comment.
A lot of the complexity where will only last for the transition period, because we need to support both the old and new way that main is exported.
I'm hoping we can simplify this again once https://reviews.llvm.org/D75277 lands.
I don't think a wrapper will work here because __main_argc_argv is the mangled name that clang will start producing for the C main. (This is done using the same mangling mechanism used for C++ symbols in clang.. so __main_argc_argv is main once https://reviews.llvm.org/D75277 lands).
That PR landed before I managed to push these two minor updates.
This change should allow https://reviews.llvm.org/D75277 to land
without breaking emscripten.
In the long run I hope to remove all of this in favor of calling
_startfrom JS, but that can happen later, and could be user-visibleif folks are calling
mainthemselves.