reexec

.

Why not just use std::env::current_exe()?

Advantages of reexec:

• Most OSes have multiple fallback methods (e.g. may work on Linux even if /proc isn't mounted)
• Sometimes able to re-execute the original program even if it has been replaced (only works on certain platforms, and only when /proc is mounted)
• Often able to avoid allocating memory
• Has an unsafe lower-level interface which some programs may find helpful

Disadvantages of reexec:

• OS support is slightly more limited (see below)

Supported OSes

Tested in CI:

• Linux (glibc and musl)
• macOS
• Windows
• FreeBSD

Verified to work on:

• NetBSD
• OpenBSD (but see below)

Untested:

• Solaris/IllumOS
• DragonFlyBSD
• Redox

OS-specific notes

OpenBSD

OpenBSD doesn't provide a straightforward way to get the executable path, so this is done by looking at the first command-line argument. Hence this will only work if argv[0] is an absolute or relative path, not just a program name.

std::env::current_exe() does this too. However, reexec actually verifies that the path specified in argv[0] refers to the current executable; std does not. (The disadvantage of this is that reexec will not be able to get the executable path if the program has been replaced.)