Rework time related code #64
Conversation
These changes begin the process of decoupling timing code from threading code, adding additional timing functionality, and removing unused functionality. * Add conversion functions between timespec and timeval * Add arithmetic functions for timespec and timeval * Make sane the public interface for internal timing functions * Remove false comment about quagga_get_relative updating recent_time on systems where gettimeofday() is used to compute a relative system time, since this never happens... * Fix bug on __APPLE__ where the argument to quagga_get_relative is not actually set with the computed system monotonic time * Decouple thread.c time caching from actual timing logic * Remove global variable recent_time because it is never used * Add FRR_CLK_REALTIME to supported clocks * Document new and existing functionality. * Rename quagga_* time functions to frr_* time functions Signed-off-by: Quentin Young <qlyoung@cumulusnetworks.com>
Signed-off-by: Quentin Young <qlyoung@cumulusnetworks.com>
* Include timeutil.h where necessary * Rename frr_* time functions to be project name agnostic * Refactor to use renamed time functions Signed-off-by: Quentin Young <qlyoung@cumulusnetworks.com>
qlyoung
force-pushed
the
time-organize
branch
from
d6ef0a6 to
6bfbc40
Compare
Continous Integration Result: SUCCESSFULCongratulations, this patch passed basic tests Tested-by: NetDEF / OpenSourceRouting.org CI System CI System Testrun URL: https://ci1.netdef.org/browse/FRR-FRRPULLREQ-35/ This is a comment from an EXPERIMENTAL automated CI system. |
Continous Integration Result: SUCCESSFULCongratulations, this patch passed basic tests Tested-by: NetDEF / OpenSourceRouting.org CI System CI System Testrun URL: https://ci1.netdef.org/browse/FRR-FRRPULLREQ-36/ This is a comment from an EXPERIMENTAL automated CI system. |
|
Mhm. I'd go in a somewhat different direction in reworking this, particularly I don't see the point of TU_CLK_REALTIME. Need to put some brain-cycles on it. I fully agree with ripping out recent_time and splitting it off thread.c though. Also, the first commit on this is marked WIP, does that mean the pull request will change further? |
|
@eqvinox and myself discussed these changes some more. We came to the conclusion that the intent is no longer to pretend to emulate various system time calls; timeutil.[ch] will now provide exactly one clock -- a monotonic clock -- and the rest of the codebase will use gettimeofday() for wall time. Per this move the various wrappers for a monotonic clock will be removed. Closing this PR as @eqvinox will submit a combined patch with these changes. |
This is a set of changes that aim to clean up timing code.
Summary of changes: