Cygwin sysconf.cc

Cygwin sysconf.cc

[Brian Inglis]

Interested in a patch for sysconf.cc to return:

      _SC_TZNAME_MAX => TZNAME_MAX and
_SC_MONOTONIC_CLOCK => _POSIX_MONOTONIC_CLOCK?

-- 
Take care. Thanks, Brian Inglis, Calgary, Alberta, Canada

This email may be disturbing to some readers as it contains
too much technical detail. Reader discretion is advised.
[Data in binary units and prefixes, physical quantities in SI.]

Re: Cygwin sysconf.cc

[Corinna Vinschen]

Hi Brian,

On Mar  1 13:20, Brian Inglis wrote:
> Interested in a patch for sysconf.cc to return:
> 
>      _SC_TZNAME_MAX => TZNAME_MAX and
> _SC_MONOTONIC_CLOCK => _POSIX_MONOTONIC_CLOCK?

not sure I understand the question.  Both are already implemented.

  $ getconf -a | egrep '(TZNAME_MAX|MONOTONIC_CLOCK)'
  _POSIX_TZNAME_MAX                   6
  TZNAME_MAX                          undefined
  _POSIX_MONOTONIC_CLOCK              200809


Corinna

Re: Cygwin sysconf.cc

[Brian Inglis]

On 2022-03-02 01:25, Corinna Vinschen wrote:
> Hi Brian,
> 
> On Mar  1 13:20, Brian Inglis wrote:
>> Interested in a patch for sysconf.cc to return:
>>
>>       _SC_TZNAME_MAX => TZNAME_MAX and
>> _SC_MONOTONIC_CLOCK => _POSIX_MONOTONIC_CLOCK?
> 
> not sure I understand the question.  Both are already implemented.
> 
>    $ getconf -a | egrep '(TZNAME_MAX|MONOTONIC_CLOCK)'
>    _POSIX_TZNAME_MAX                   6
>    TZNAME_MAX                          undefined
>    _POSIX_MONOTONIC_CLOCK              200809

Sorry, must have been looking at very *OLD* version online, as 
_SC_CLOCK_SELECTION and _SC_MONOTONIC_CLOCK were not defined.

Why did you not define _SC_TZNAME_MAX => _POSIX_TZNAME_MAX when you 
tweaked it?

My rereading of the man and POSIX pages leads me to believe that for all 
known values of _SC_... the entries now showing {nsup, {c:0}} should be 
{cons, {c:-1L}} supported but undefined, and only out of range values 
for the parameter should be treated as {nsup, {c:-1L}}?

-- 
Take care. Thanks, Brian Inglis, Calgary, Alberta, Canada

This email may be disturbing to some readers as it contains
too much technical detail. Reader discretion is advised.
[Data in binary units and prefixes, physical quantities in SI.]

Re: Cygwin sysconf.cc

[Corinna Vinschen]

On Mar  2 12:45, Brian Inglis wrote:
> On 2022-03-02 01:25, Corinna Vinschen wrote:
> > Hi Brian,
> > 
> > On Mar  1 13:20, Brian Inglis wrote:
> > > Interested in a patch for sysconf.cc to return:
> > > 
> > >       _SC_TZNAME_MAX => TZNAME_MAX and
> > > _SC_MONOTONIC_CLOCK => _POSIX_MONOTONIC_CLOCK?
> > 
> > not sure I understand the question.  Both are already implemented.
> > 
> >    $ getconf -a | egrep '(TZNAME_MAX|MONOTONIC_CLOCK)'
> >    _POSIX_TZNAME_MAX                   6
> >    TZNAME_MAX                          undefined
> >    _POSIX_MONOTONIC_CLOCK              200809
> 
> Sorry, must have been looking at very *OLD* version online, as
> _SC_CLOCK_SELECTION and _SC_MONOTONIC_CLOCK were not defined.
> 
> Why did you not define _SC_TZNAME_MAX => _POSIX_TZNAME_MAX when you tweaked
> it?

Because it's wrong.  _POSIX_TZNAME_MAX is just a minimum value required
by POSIX, not the correct value to return for TZNAME_MAX.

> My rereading of the man and POSIX pages leads me to believe that for all
> known values of _SC_... the entries now showing {nsup, {c:0}} should be
> {cons, {c:-1L}} supported but undefined, and only out of range values for
> the parameter should be treated as {nsup, {c:-1L}}?

These are really not undefined, but not supported on Cygwin.  That's
why they return with EINVAL.  I see what you mean, though, let me think
about it.

while looking into this I see at least one obvious bug.  While adding
POSIX per-process timers in 2019 I added a valid DELAYTIMER_MAX value,
but I neglected to add this to sysconf.  I'm going to fix this.


Corinna

Re: Cygwin sysconf.cc

[Corinna Vinschen]

On Mar  2 21:35, Corinna Vinschen wrote:
> On Mar  2 12:45, Brian Inglis wrote:
> > On 2022-03-02 01:25, Corinna Vinschen wrote:
> > > Hi Brian,
> > > 
> > > On Mar  1 13:20, Brian Inglis wrote:
> > > > Interested in a patch for sysconf.cc to return:
> > > > 
> > > >       _SC_TZNAME_MAX => TZNAME_MAX and
> > > > _SC_MONOTONIC_CLOCK => _POSIX_MONOTONIC_CLOCK?
> > > 
> > > not sure I understand the question.  Both are already implemented.
> > > 
> > >    $ getconf -a | egrep '(TZNAME_MAX|MONOTONIC_CLOCK)'
> > >    _POSIX_TZNAME_MAX                   6
> > >    TZNAME_MAX                          undefined
> > >    _POSIX_MONOTONIC_CLOCK              200809
> > 
> > Sorry, must have been looking at very *OLD* version online, as
> > _SC_CLOCK_SELECTION and _SC_MONOTONIC_CLOCK were not defined.
> > 
> > Why did you not define _SC_TZNAME_MAX => _POSIX_TZNAME_MAX when you tweaked
> > it?
> 
> Because it's wrong.  _POSIX_TZNAME_MAX is just a minimum value required
> by POSIX, not the correct value to return for TZNAME_MAX.
> 
> > My rereading of the man and POSIX pages leads me to believe that for all
> > known values of _SC_... the entries now showing {nsup, {c:0}} should be
> > {cons, {c:-1L}} supported but undefined, and only out of range values for
> > the parameter should be treated as {nsup, {c:-1L}}?
> 
> These are really not undefined, but not supported on Cygwin.  That's
> why they return with EINVAL.  I see what you mean, though, let me think
> about it.

Yep, I guess you're right.  I compared this with what Linux returns
for the unsupported tracing options.  See commits cf00bba99a61 and
fcec4830abf0.


Thanks,
Corinna