<lsb-release> reader #1781

[surahman]

With respect to issue #1781:
I have developed tests and utility routines to read the lsb-release file on Ubuntu-based systems when applicable. When it is unavailable, or parsing fails, there will be a failover to the previous method of using SysInfo::productType() and QSysInfo::productVersion().

[ricab]

Hi @surahman, thanks a lot for your work on this!

Unfortunately, we've discussed the disparity between /etc/lsb-release and /etc/os-release and it's not looking like we can rely on any of those. The fact that they're readable at all is misleading.

Instead, we should probably read from /var/lib/snapd/hostfs/etc/os-release and add the system-observe plug to our snap. We'd then have to decide what to do if the user doesn't give us the plug.

We're still discussing details, and we'll get back to you again, but we wanted to let you know so that you don't spend any more time.

[ricab]

Hi again @surahman,

So we discussed a bit and confirmed /var/lib/snapd/hostfs/etc/os-release is the way to go, given how snaps are working... The implementation can just return "unknown" when unable to read that file (that's what QSysInfo defaults to as well).

Multipass needs to work on macOS/Windows too, so this would need to be a platform dependent implementation. That can be achieved by declaring a free function in include/multipass/platform.h and implementing it in src/platform/platform_linux.cpp (we'd take care of other platforms). The code you now have for parsing lsb-release would need to be adapted and moved to that compilation unit. The corresponding tests go in tests/linux/test_platform_linux.cpp.

Notice that the Linux implementation should still use QSysInfo when outide of a snap. So, something like

std::string multipass::platform::host_version() // returning single string probably enough for our purposes
{
  return mu::in_multipass_snap() ? 
    our_snap_impl_reading_os_release() :
    fmt::format("{}-{}", QSysInfo::productType(), QSysInfo::productVersion());
}

Let us know if you have questions and thanks again for your involvement!

[surahman]

I am on it. I am happy to help and I am learning a lot from your codebase. These interactions and exercises are a great learning/practice tool and it is even better if it is of assistance to someone.

[surahman]

I will go ahead and abandon this branch and PR to start on a new branch - it is the cleanest way to revert changes. I will cross-reference the two PR's once I have the new one up.

[surahman]

I have created a new branch instead of doing a hard reset on this one: #2105

[Saviq]

Thanks @surahman, I'll close this one then.