Bugzilla

Now that i'm rereading the manpages and freebsd's code and openbsd's code, i understand things aren't so simple, since Linux, OpenBSD and FreeBSD don't seem to agree on the BOOTTIME/UPTIME meaning.

FreeBSD's manpage says
> Finally, CLOCK_BOOTTIME is an alias for CLOCK_UPTIME for compatibility with other systems.

while for OpenBSD's, the code and the manpage seems to mean that CLOCK_UPTIME should be used in `NowExcludingSuspendMs` and CLOCK_BOOTIME in `
NowIncludingSuspendMs`. The more i think about it, the more it seems https://phabricator.services.mozilla.com/D101748 was right.

to check the values, in about:telemetry i'm checking `browser.engagement.session_time_including_suspend` and `browser.engagement.session_time_excluding_suspend`  in the scalars/parent tree. With my current fix (adding `|| defined (XP_OPENBSD)`, i have the same values for both, even after a suspend/resume session.

Will recheck with an adaptation/update of the aforementioned bitrotten patch. Val, did you end up with the same analysis as me ?

Of course, just replacing XP_LINUX by XP_UNIX wont break build for any BSD platform and will enable telemetry, just with a single value not reporting 'correct' information for OpenBSD. It's a simpler patch, but should we strive for absolute correctness and do the #ifdef dance ?

Now that i'm rereading the manpages and freebsd's code and openbsd's code, i understand things aren't so simple, since Linux, OpenBSD and FreeBSD don't seem to agree on the BOOTTIME/UPTIME meaning.

FreeBSD's manpage says
> Finally, CLOCK_BOOTTIME is an alias for CLOCK_UPTIME for compatibility with other systems.

while for OpenBSD's, the code in https://github.com/openbsd/src/blob/50603113d13b32b82b4569f48902052db016a195/sys/kern/kern_time.c#L119 and the manpage seems to mean that CLOCK_UPTIME should be used in `NowExcludingSuspendMs` and CLOCK_BOOTIME in `
NowIncludingSuspendMs`. The more i think about it, the more it seems https://phabricator.services.mozilla.com/D101748 was right.

to check the values, in about:telemetry i'm checking `browser.engagement.session_time_including_suspend` and `browser.engagement.session_time_excluding_suspend`  in the scalars/parent tree. With my current fix (adding `|| defined (XP_OPENBSD)`, i have the same values for both, even after a suspend/resume session.

Will recheck with an adaptation/update of the aforementioned bitrotten patch. Val, did you end up with the same analysis as me ?

Of course, just replacing XP_LINUX by XP_UNIX wont break build for any BSD platform and will enable telemetry, just with a single value not reporting 'correct' information for OpenBSD. It's a simpler patch, but should we strive for absolute correctness and do the #ifdef dance ?

Now that i'm rereading the manpages and freebsd's code and openbsd's code, i understand things aren't so simple, since Linux, OpenBSD and FreeBSD don't seem to agree on the BOOTTIME/UPTIME meaning.

FreeBSD's manpage says
> Finally, CLOCK_BOOTTIME is an alias for CLOCK_UPTIME for compatibility with other systems.

while for OpenBSD's, the code in https://github.com/openbsd/src/blob/50603113d13b32b82b4569f48902052db016a195/sys/kern/kern_time.c#L119 and the manpage seems to mean that CLOCK_UPTIME should be used in `NowExcludingSuspendMs` and CLOCK_BOOTIME in `
NowIncludingSuspendMs`. The more i think about it, the more it seems https://phabricator.services.mozilla.com/D101748 was right.

to check the values, in about:telemetry i'm checking `browser.engagement.session_time_including_suspend` and `browser.engagement.session_time_excluding_suspend`  in the scalars/parent tree. (eg `about:telemetry#scalars-tab_search=sus`) With my current fix (adding `|| defined (XP_OPENBSD)`, i have the same values for both, even after a suspend/resume session.

Will recheck with an adaptation/update of the aforementioned bitrotten patch. Val, did you end up with the same analysis as me ?

Of course, just replacing XP_LINUX by XP_UNIX wont break build for any BSD platform and will enable telemetry, just with a single value not reporting 'correct' information for OpenBSD. It's a simpler patch, but should we strive for absolute correctness and do the #ifdef dance ?