client (Unix): add support for idle detection in a separate process

[davidpanderson]

The design is described here:
https://github.com/BOINC/boinc/wiki/Linux-idle-detection

[Copilot]

Copilot reviewed 2 out of 2 changed files in this pull request and generated 2 comments.

[Copilot]

Typo in the comment: 'permssions' should be corrected to 'permissions'.

Suggested change

	// (big mess, permssions problems, doesn't generally work)
	// new approach: get last-input time from a separate daemon process,
	// (big mess, permissions problems, doesn't generally work)

[Copilot]

[nitpick] Consider using the project's logging mechanism instead of printf for production logging to ensure consistency and proper log level management.

[jamescowens]

@davidpanderson do you mind making these signed 64 bit ints instead of unsigned so that we conform to the new time standard?

POSIX requires time to be a signed int. The complete spec in <time.h> is actually

struct timespec {
    time_t   tv_sec;        /* seconds */
    long     tv_nsec;       /* nanoseconds */
};

Where time_t is typedefd internally in the header file to a 64 bit signed integer type on that platform.

We obviously don't need the nanoseconds part for this so just using the 64 bit signed integer is perfectly fine.

[davidpanderson]

Uh, sure.

[AenBleidd]

This PR fails during the build for Android:

hostinfo_unix.cpp:567:52: warning: comparison of integers of different signs: 'int' and 'unsigned int' [-Wsign-compare]
            strlcpy(buf2, strchr(buf, ':') + 2, ((t<sizeof(buf2))?t:sizeof(buf2)));
                                                  ~^~~~~~~~~~~~~
hostinfo_unix.cpp:2219:18: error: use of undeclared identifier 'shm_open'
        int fd = shm_open("/idle_detect_shmem",  O_RDONLY, 0);
                 ^
1 warning and 1 error generated.

[davidpanderson]

fixed (I hope)