[libc][POSIX] implement system()

[Wrench56]

You can find the description here

I will start working on it.

[llvmbot]

@llvm/issue-subscribers-libc

Author: Mark Devenyi (Wrench56)

You can find the description [here](https://man7.org/linux/man-pages/man3/system.3.html)

I will start working on it.

[nickdesaulniers]

I highly expect this to be system specific, so expect to put the implementation under a linux/ dir (at least to start with, assuming you're developing on Linux).

[Wrench56]

@nickdesaulniers Yes, absolutely, it is platform specific. To be honest, it should be under a POSIX folder (since I would be using POSIX threads) but since there is no such directory under stdlib, I will just put it in stdlib/linux for now.

[Wrench56]

Okay, just for the record:

The 2024 POSIX standard describes system() to mimic fork() + exec*() behaviour. Obviously, calling fork() is not memory efficient, since the address space of the parent process is copied. I'm considering implementing it with posix_spawn() instead to NOT copy the memory address space at all. Of course, I would need to still implement the SIGINT and SIGQUIT ignore and consider the other specifications. But as for the behaviour of posix_spawn(), does anybody see a reason NOT to do this?

Note: vfork() similarly is a better option that stays even closer to fork() while sharing the address space. If posix_spawn() deviates at all from fork(), I would consider vfork() next.

[amemov]

@Wrench56 are you still working on this issue?

[Wrench56]

Hey @amemov !

At the time when I opened the issue, many of the libc prerequisites were missing for writing something as "high-level" as system(). It has been a cumbersome process to refactor the existing functions to be usable from internal infrastructure and I gave up when I stumbled across entirely missing implementations. If something changed eversince, I would gladly finish up the work I started. If nothing changed, I'm afraid my hands are tied. If you feel ambitious, you are more than welcome to pick of my fork where I left it: Wrench56@1d8cf5f.

Let me know, how you decide.
Cheers!

[amemov]

@Wrench56 , thanks for the response and sharing what you wrote! I'm not sure if I will attempt to take over this issue yet, considering that some functions are still missing.

I'm currently looking for next issue to solve in the Libc that is more related to OS after I finalize my PR in the for the Libc's Math library. But, I don't want to spread too thin right now, so I will try to revisit this issue probably in ~2 weeks ( or a little later ) to see what is missing as of right now, and from where I will start working towards finalizing this implementation. I will update you here when I will start working towards resolving this issue!