How does libc provide functions with two names?
Before the advent of direct binding (-B direct) libc provided many functions with two names. For example, getpwent() and _getpwent(). These two names referred to exactly the same function in libc.
How does libc开发者_JAVA百科 make two function names point to the same implementation?
I think it should not be as easy as writing the same code twice though.
It's done via weak aliases a "nonstandard" linker trick that's been around since early unices and that's supported by all unix compilers/linkers I know of. It's basically done as:
void __foo(void);
void foo(void) __attribute__((weak, alias("__foo")));
often with macros to abstract it a little bit. This makes it so the symbol foo
will have the same address and type as the symbol __foo
by default, but allows it to be overridden by a "strong" definition somewhere else.
getpwent() implementation just calls _getpwent() simple as that. The reason this is done is to hide some functionality from function calls and to avoid something called namespace pollution. This way you can create a sort of abstraction that allows you to hide things from the user. Also leading underscore and double underscore are system reserved and are backups to make sure that you don't override something such as in macro definitions.
精彩评论