Math error in libbsd timespecsub

[Reggie McMurtrey]

#define timespecsub(tsp, usp, vsp) \
        do { \
                (vsp)->tv_sec = (tsp)->tv_sec - (usp)->tv_sec; \
                (vsp)->tv_nsec = (tsp)->tv_nsec - (usp)->tv_nsec; \
                if ((vsp)->tv_nsec < 0) { \
                        (vsp)->tv_sec--; \
                        (vsp)->tv_nsec += 1000000000L; \
                } \
        } while (0)
#endif
*******************************************
the line

"if ((vsp)->tv_nsec < 0) {"

should be

"if ((vsp)->tv_nsec < 0 && (vsp)->tv_sec != 0) {"

try tsp.tv_sec = 0, tsp.tv_nsec = 1, usp.tv_sec = 0, usp.tv_nsec = 2 as a test case

I found this trying to calculate how long to nanosleep for.

[Thorsten Glaser (mirabilos)]

No, the code is correct.

If you calculate (for the sake of simplicity I’m skipping the 0s) 0.1 - 0.2 you get -0.1 which in timespec is the same as -1 + .9.

If you’re trying to calculate how long to sleep for, you probably shouldn’t sleep *at all* if your destination timestamp is already in the past.

[Reggie McMurtrey (reggie-mcmurtrey)]

Ok, I see what you are doing and why, but it may be worth putting a blurb
in the timespecsub man page about how you are handling negative numbers.
I'm fighting another bug in a system where we aren't meeting our deadline
(why we are gettiing negative sleep #'s). we added some debug code to see
what our sleep values were and the values didn't make sense.

On Fri, Aug 25, 2023 at 5:30 PM Thorsten Glaser <email address hidden>
wrote:

> No, the code is correct.
>
> If you calculate (for the sake of simplicity I’m skipping the 0s) 0.1 -
> 0.2 you get -0.1 which in timespec is the same as -1 + .9.
>
> If you’re trying to calculate how long to sleep for, you probably
> shouldn’t sleep *at all* if your destination timestamp is already in the
> past.
>
> --
> You received this bug notification because you are subscribed to the bug
> report.
> https://bugs.launchpad.net/bugs/2033113
>
> Title:
> Math error in libbsd timespecsub
>
> Status in libbsd package in Ubuntu:
> New
>
> Bug description:
> #define timespecsub(tsp, usp, vsp) \
> do { \
> (vsp)->tv_sec = (tsp)->tv_sec - (usp)->tv_sec; \
> (vsp)->tv_nsec = (tsp)->tv_nsec - (usp)->tv_nsec; \
> if ((vsp)->tv_nsec < 0) { \
> (vsp)->tv_sec--; \
> (vsp)->tv_nsec += 1000000000L; \
> } \
> } while (0)
> #endif
> *******************************************
> the line
>
> "if ((vsp)->tv_nsec < 0) {"
>
> should be
>
> "if ((vsp)->tv_nsec < 0 && (vsp)->tv_sec != 0) {"
>
> try tsp.tv_sec = 0, tsp.tv_nsec = 1, usp.tv_sec = 0, usp.tv_nsec = 2
> as a test case
>
> I found this trying to calculate how long to nanosleep for.
>
> To manage notifications about this bug go to:
>
> https://bugs.launchpad.net/ubuntu/+source/libbsd/+bug/2033113/+subscriptions
>
>