Paths

Table of Contentst

-
sys/sys/
-
sys/
7/12
endian.h

Differential D31964

endian.h: Use the __bswap* versions
ClosedPublic
Actions

Authored by imp on Sep 14 2021, 10:23 PM.

Details

Reviewers

emaste
jhb
markj

Commits

rG6f3e6234f8be: endian.h: Use the __bswap* versions
rG44fb3c695f5e: endian.h: Use the __bswap* versions

Summary

Make it possible to have all these macros work without bswap* being
defined. bswap* is part of the application namespace and applications
are free to redefine those functions.

Diff Detail

Repository

rG FreeBSD src repository

Lint

Lint Not Applicable

Unit

Tests Not Applicable

Event Timeline

imp requested review of this revision.Sep 14 2021, 10:23 PM

imp created this revision.

Harbormaster completed remote builds in B41518: Diff 95183.Sep 14 2021, 10:23 PM

imp edited the test plan for this revision. (Show Details)Sep 14 2021, 10:24 PM

imp added reviewers: emaste, jhb, markj.

imp added a child revision: D31962: linux: For better compatibility, provide compatible endian.h.

markj accepted this revision.Sep 15 2021, 1:23 PM

This revision is now accepted and ready to land.Sep 15 2021, 1:23 PM

emaste accepted this revision.Sep 15 2021, 1:40 PM

jhb accepted this revision.Sep 15 2021, 5:25 PM

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=258544 just to be sure...

imp mentioned this in D19500: Add endian.h to LHDRS..Sep 16 2021, 7:57 PM

After the exp run, this likely is fine to go in. No failures from this change were detected.

Closed by commit rG44fb3c695f5e: endian.h: Use the __bswap* versions (authored by imp). · Explain WhySep 21 2021, 4:10 AM

This revision was automatically updated to reflect the committed changes.

imp added a commit: rG44fb3c695f5e: endian.h: Use the __bswap* versions.

imp added a commit: rG6f3e6234f8be: endian.h: Use the __bswap* versions.Jan 24 2023, 11:28 PM

Revision Contents
Changeset List

Path

Size

sys/

endian.h

24 lines

Diff 95420

View Options

sys/sys/endian.h

Show All 26 Lines

.\" HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT .\" HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT

.\" LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY .\" LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY

.\" OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF .\" OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF

.\" SUCH DAMAGE. .\" SUCH DAMAGE.

.\" .\"

.\" @(#)time.3 8.1 (Berkeley) 6/4/93 .\" @(#)time.3 8.1 (Berkeley) 6/4/93

.\" $FreeBSD$ .\" $FreeBSD$

.\" .\"

.Dd March 4, 2022 .Dd April 2, 2022

.Dt TIME 3 .Dt TIME 3

.Os .Os

.Sh NAME .Sh NAME

.Nm time .Nm time

.Nd get time of day .Nd get time of day

.Sh LIBRARY .Sh LIBRARY

.Lb libc .Lb libc

.Sh SYNOPSIS .Sh SYNOPSIS

.In time.h .In time.h

.Ft time_t .Ft time_t

.Fn time "time_t *tloc" .Fn time "time_t *tloc"

.Sh DESCRIPTION .Sh DESCRIPTION

The The

.Fn time .Fn time

function function

returns the value of time in seconds since 0 hours, 0 minutes, returns the value of time in seconds since 0 hours, 0 minutes,

0 seconds, January 1, 1970, Coordinated Universal Time. 0 seconds, January 1, 1970, Coordinated Universal Timei (UTC).

impUnsubmitted

Done

extra 'i' here.

imp: extra 'i' here.

If an error occurs, If an error occurs,

.Fn time .Fn time

returns the value returns the value

.Po Vt time_t Pc Ns \-1 . .Po Vt time_t Pc Ns \-1 .

.Pp .Pp

The return value is also stored in The return value is also stored in

.No \&* Ns Va tloc , .No \&* Ns Va tloc ,

provided that provided that

.Va tloc .Va tloc

is non-null. is non-null.

.Sh ERRORS .Sh ERRORS

The The

.Fn time .Fn time

function may fail for any of the reasons described in function may fail for any of the reasons described in

.Xr clock_gettime 2 . .Xr clock_gettime 2 .

.Sh SEE ALSO .Sh SEE ALSO

.Xr clock_gettime 2 , .Xr clock_gettime 2 ,

.Xr gettimeofday 2 , .Xr gettimeofday 2 ,

.Xr ctime 3 .Xr ctime 3

.Sh STANDARDS .Sh STANDARDS

The The

.Nm .Nm

function conforms to function conforms to

.St -p1003.1-2001 . .St -p1003.1-2008 .

.Sh HISTORY .Sh HISTORY

A A

.Fn time .Fn time

function appeared in system call first appeared in

.At v6 . .At v1 .

That version counted time in sixtieths of a second with a 32-bit return value,

pauamma_gundo.comUnsubmitted

Not Done

.At v1 .

- Through the

- .At v3 ,

+ Until and including.At v3 ,

pauamma_gundo.com:

ensuring an integer overflow crisis every 2.26 years.

.At v6

impUnsubmitted

Done

This change to seconds happened in the 4th edition, not the 6th edition.
The 3rd edition returned time in 1/HZ (1/60) since Jan 1, 1972.
The 1st and 2nd Edition return since Jan 1, 1971

imp: This change to seconds happened in the 4th edition, not the 6th edition. The 3rd edition…

impUnsubmitted

Done

The new epoch in the 3rd edition presents challenges to reconstructing dates from tapes...
And the headaches with it motivated the change in the 4th edition...

imp: The new epoch in the 3rd edition presents challenges to reconstructing dates from tapes... And…

the granularity of the return value was reduced to whole seconds,

delaying the aforementioned crisis until 2038.

.Bx 4.1c

impUnsubmitted

Done

This is close, but not quite right.

v7 introduced the ftime() system call which returned time at a millisecond level, though retained the gtime() system call (exposed as time() in userland). time() could have been implemented as a wrapper around ftime, but that wasn't done.

4.1c implemented a higher-precision time function (gettimeofday) to replace ftime() and reimplemented time() in terms of that.

imp: This is close, but not quite right. v7 introduced the ftime() system call which returned time…

jillesUnsubmitted

Done

And a more recent change: git commit 358ed16f750576f266ebf86ea1ee56204d71bb52 (in 9.0 and 8.1) changed the implementation of time() to use clock_gettime(CLOCK_SECOND) instead of gettimeofday(), for performance reasons.

The practical effect is that applications should use time() if a resolution of 1 second is good enough for them; time() is standard but CLOCK_SECOND is not.

jilles: And a more recent change: git commit 358ed16f750576f266ebf86ea1ee56204d71bb52 (in 9.0 and 8.1)…

the function was moved out of the kernel into the C standard library and

impUnsubmitted

Done

I'd make this briefer:
"The time system call first appeared in 1st edition. Through the 3rd edition, it returned 60 Hz ticks since an epoch that changed occasionally because it was a 32-bit value that overflowed in a little over 2 years."
and maybe leave out the tape date information. It's a bit more verbose than we normally do. And tape dates varied because they were both the 1971 and 1972 epoch, and you never knew which type of tape you had.

imp: I'd make this briefer: "The time system call first appeared in 1st edition. Through the 3rd…

reimplemented with

.Xr gettimeofday 2 .

.Sh BUGS .Sh BUGS

Neither Neither

.St -isoC-99 .St -isoC-99

nor nor

pauamma_gundo.comUnsubmitted

Not Done

system call, which returned time at a millisecond level,

- though retained the

+ though it retained the

.Fn gtime

pauamma_gundo.com:

.St -p1003.1-2001 .St -p1003.1-2001

requires requires

impUnsubmitted

Not Done

i386 is the only platform left with a 32-bit time_t, and the only one affected by the crisis. not entirely sure how to work that into this, however.

imp: i386 is the only platform left with a 32-bit time_t, and the only one affected by the crisis.

gbeAuthorUnsubmitted

Done

I left this out for the moment, since I am not sure what about arm6, or is i386 the only left 32bit architecture?

gbe: I left this out for the moment, since I am not sure what about arm6, or is i386 the only left…

pauamma_gundo.comUnsubmitted

Not Done

I left this out for the moment, since I am not sure what about arm6, or is i386 the only left 32bit architecture?

https://cgit.freebsd.org/src/tree/sys/arm/include/_types.h (which I think is arm[4-6]) has

typedef	__int64_t	__time_t;		/* time()... */

pauamma_gundo.com: > I left this out for the moment, since I am not sure what about arm6, or is i386 the only left…

.Fn time .Fn time

to set to set

.Va errno .Va errno

on failure; thus, it is impossible for an application to distinguish on failure; thus, it is impossible for an application to distinguish

the valid time value \-1 (representing the last UTC second of 1969) the valid time value \-1 (representing the last UTC second of 1969)

from the error return value. from the error return value.

.Pp .Pp

Systems conforming to earlier versions of the C and Systems conforming to earlier versions of the C and

.Tn POSIX .Tn POSIX

standards (including older versions of standards (including older versions of

.Fx ) .Fx )

did not set did not set

.No \&* Ns Va tloc .No \&* Ns Va tloc

in the error case. in the error case.

pauamma_gundo.comUnsubmitted

Not Done

.Fn time

- in terms of that.

+ using that.

.Pp

Since

pauamma_gundo.com:

endian.h: Use the __bswap* versionsClosedPublicActions

Details

Diff Detail

Event Timeline

Revision ContentsChangeset List

Diff 95420

sys/sys/endian.h

endian.h: Use the __bswap* versions
ClosedPublic
Actions

Revision Contents
Changeset List