Re: [HACKERS] Problem with copying abstimes - Mailing list pgsql-hackers

- --NeXT-Mail-1349238861-1
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: quoted-printable

You wrote:
> > > AIX 3.2.5 supports neither int timezone (even tho configure seems =
to
> > > think it does) nor the Sun/DEC tm struct with tm_gmtoff field. The
> > > non-POSIX time handling seems just plain broken. My patches will
> > > therefore only affect AIX 3.2.5 and NeXTstep, the other platform
> > > without POSIX_TIME. AIX 4.1 supports POSIX_TIME and is therefore =
not
> > > a problem.
>=20
> I would think that NextStep is at risk of getting dropped from =
Postgres
> support, since it does not have shared memory or, apparently, anything
> close to POSIX time support. If there is someone running on NextStep, =
it
> would be good if they would make themselves known so we can see what =
the
> NextStep OS support needs to be (can't believe it _still_ doesn't have
> anything like POSIX-compatible time support).

Beep. Wait a moment...

    hactar:6> uname -a
    NEXTSTEP hactar Rose2L 3.3 (7100LC) HPPA

Accepted ?

NEXTSTEP indeed has no SysV IPC and shared memory interface, but there's =
a commercial third-party product, SysVIPC (free for academic use), that =
implements those routines based on the Mach shared memory interface. =
That's enough to compile and run Postgres95 (in fact the lib was written =
for this purpose). Maybe one should add this to the docs (should I write =
down a paragraph ?).

NEXTSTEP comes with a  POSIX compatibility library, but this is heavily =
broken so that everybody recommends against using it, therefore one is =
stuck with plain the BSD 4.3 interfaces.

Regarding USE_POSIX_TIME: =46rom browsing through dt.c, datetime.c and =
nabstime.c, it looks like NEXTSTEP's ctime(3) has all you need for =
USE_POSIX_TIME. Since the NEXTSTEP port is not functional at all at the =
moment, I would suggest to switch to USE_POSIX_TIME for NEXTSTEP and =
drop the old stuff. If that should break time support for NEXTSTEP, we =
have to look at it later again.

Ok ?

    Gregor


- ---
| Gregor Hoffleit                        Mathematisches Institut, Uni HD =
|
| flight@mathi.uni-heidelberg.de      INF 288, 69120 Heidelberg, Germany =
|
| (NeXTmail, MIME)                      (49)6221 54-5771    fax  54-8312 =
|
| PGP Key fingerprint =3D 23 8F B3 38 A3 39 A6 01  5B 99 91 D6 F2 AC CD =
C7 |

- --NeXT-Mail-1349238861-1
Content-Type: text/enriched; charset=us-ascii
Content-Transfer-Encoding: quoted-printable

You wrote:

> > > AIX 3.2.5 supports neither int timezone (even tho configure seems =
to

> > > think it does) nor the Sun/DEC tm struct with tm_gmtoff field. The

> > > non-POSIX time handling seems just plain broken. My patches will

> > > therefore only affect AIX 3.2.5 and NeXTstep, the other platform

> > > without POSIX_TIME. AIX 4.1 supports POSIX_TIME and is therefore =
not

> > > a problem.

>=20

> I would think that NextStep is at risk of getting dropped from =
Postgres

> support, since it does not have shared memory or, apparently, anything

> close to POSIX time support. If there is someone running on NextStep, =
it

> would be good if they would make themselves known so we can see what =
the

> NextStep OS support needs to be (can't believe it _still_ doesn't have

> anything like POSIX-compatible time support).


Beep. Wait a moment...


    hactar:6> uname -a

    NEXTSTEP hactar Rose2L 3.3 (7100LC) HPPA


Accepted ?


NEXTSTEP indeed has no SysV IPC and shared memory interface, but there's =
a commercial third-party product, SysVIPC (free for academic use), that =
implements those routines based on the Mach shared memory interface. =
That's enough to compile and run Postgres95 (in fact the lib was written =
for this purpose). Maybe one should add this to the docs (should I write =
down a paragraph ?).


NEXTSTEP comes with a  POSIX compatibility library, but this is heavily =
broken so that everybody recommends against using it, therefore one is =
stuck with plain the BSD 4.3 interfaces.


Regarding USE_POSIX_TIME: =46rom browsing through dt.c, datetime.c and =
nabstime.c, it looks like NEXTSTEP's ctime(3) has all you need for =
USE_POSIX_TIME. Since the NEXTSTEP port is not functional at all at the =
moment, I would suggest to switch to USE_POSIX_TIME for NEXTSTEP and =
drop the old stuff. If that should break time support for NEXTSTEP, we =
have to look at it later again.


Ok ?


    Gregor



- ---

|
<smaller> Gregor Hoffleit                        Mathematisches =
Institut, Uni HD |

| flight@mathi.uni-heidelberg.de      INF 288, 69120 Heidelberg, Germany =
|

| (NeXTmail, MIME)                      (49)6221 54-5771    fax  54-8312 =
|

| PGP Key fingerprint =3D 23 8F B3 38 A3 39 A6 01  5B 99 91 D6 F2 AC CD =
C7 |

</smaller>

- --NeXT-Mail-1349238861-1--

------------------------------