[9fans] 9atom install error

[9fans] 9atom install error

[Ramakrishnan Muthukrishnan]

Erik,

I noticed this on a fresh install from the usbinstamd64.bz2 downloaded
today. I am typing the error. The error is while it is compiling the
amd64 programs..

[...]
mk syscall
sed '/#define...
....
6c -I/sys/src/libc/9syscall -FTVw syscall.c
tab.h:52 syscall.c:34 name not declared: _nsec
mk: 6c -I/sys ... : exit status=rc 20800: 6c 20802: error
mk: for(i in 1a ... : exit status=rc 9085: rc 20789: mk 20792: error
mk: test -e ... error
mk: date for (i ... : ... error
halt system? ...

I don't remember getting this error when I installed it from an
earlier installer (about a month ago, I think).

--
  Ramakrishnan

Re: [9fans] 9atom install error

[erik quanstrom]

> I noticed this on a fresh install from the usbinstamd64.bz2 downloaded
> today. I am typing the error. The error is while it is compiling the
> amd64 programs..
>
> [...]
> mk syscall
> sed '/#define...
> ....
> 6c -I/sys/src/libc/9syscall -FTVw syscall.c
> tab.h:52 syscall.c:34 name not declared: _nsec
> mk: 6c -I/sys ... : exit status=rc 20800: 6c 20802: error
> mk: for(i in 1a ... : exit status=rc 9085: rc 20789: mk 20792: error
> mk: test -e ... error
> mk: date for (i ... : ... error
> halt system? ...

ok.  there must have been version skew.  i rebuilt everything.  but i'm traveling, so i didn't
do any sanity checking.

_nsec is just for binary compatability with the nsec system call.  i'm still
unconvinced, but i don't want to introduce an incompatability like that either.

by the way, one of the least attractive things about an nsec system call is
high frequency high precision timing would be much better off with a
relative clock that is not adjusted.  cf. the linux leap second fiasco.
TSC is probablly ideal for this on intel machines.

- erik

Re: [9fans] 9atom install error

[Ramakrishnan Muthukrishnan]

On Thu, Jun 26, 2014 at 8:57 AM, erik quanstrom <quanstro@quanstro.net> wrote:
>> I noticed this on a fresh install from the usbinstamd64.bz2 downloaded
>> today. I am typing the error. The error is while it is compiling the
>> amd64 programs..
>>
>> [...]
>> mk syscall
>> sed '/#define...
>> ....
>> 6c -I/sys/src/libc/9syscall -FTVw syscall.c
>> tab.h:52 syscall.c:34 name not declared: _nsec
>> mk: 6c -I/sys ... : exit status=rc 20800: 6c 20802: error
>> mk: for(i in 1a ... : exit status=rc 9085: rc 20789: mk 20792: error
>> mk: test -e ... error
>> mk: date for (i ... : ... error
>> halt system? ...
>
> ok.  there must have been version skew.  i rebuilt everything.  but i'm traveling, so i didn't
> do any sanity checking.

I downloaded a fresh usbinstamd64.bz2 today but am getting the same
results. Don't bother, I have a virtual machine install that I can
play with.

> _nsec is just for binary compatability with the nsec system call.  i'm still
> unconvinced, but i don't want to introduce an incompatability like that either.
>
> by the way, one of the least attractive things about an nsec system call is
> high frequency high precision timing would be much better off with a
> relative clock that is not adjusted.  cf. the linux leap second fiasco.
> TSC is probablly ideal for this on intel machines.
>
> - erik
>