• Re: VMS previous DEC/CPQ/HP[E] decisions and paths

    From jgd@jgd@cix.co.uk (John Dallman) to comp.os.vms on Sun Oct 4 15:26:40 2026
    From Newsgroup: comp.os.vms

    In article <memo.20251007223453.10624Y@jgd.cix.co.uk>, jgd@cix.co.uk
    (John Dallman) wrote:

    but if Windows is only using A64 it doesn't matter.

    Microsoft supply compilers that can target 32-bit code, and run-time libraries for 32-bit programs. I've never tried building anything
    on ARM Windows for 32-bit so I don't know how well they work. I
    don't know if ARM Windows 11, which is always a 64-bit OS, will
    notice that the hardware is incapable of running A32/T32 ...

    ARM Windows 11 is not interested in running 32-bit executables and
    refuses on looking at them. I don't have any ARM Windows 10 left to try
    them on.



    John
    --- Synchronet 3.22a-Linux NewsLink 1.2
  • From Simon Clubley@clubley@remove_me.eisner.decus.org-Earth.UFP to comp.os.vms on Mon Oct 5 12:25:29 2026
    From Newsgroup: comp.os.vms

    On 2026-10-04, John Dallman <jgd@cix.co.uk> wrote:
    In article <memo.20251007223453.10624Y@jgd.cix.co.uk>, jgd@cix.co.uk
    (John Dallman) wrote:

    but if Windows is only using A64 it doesn't matter.

    Microsoft supply compilers that can target 32-bit code, and run-time
    libraries for 32-bit programs. I've never tried building anything
    on ARM Windows for 32-bit so I don't know how well they work. I
    don't know if ARM Windows 11, which is always a 64-bit OS, will
    notice that the hardware is incapable of running A32/T32 ...

    ARM Windows 11 is not interested in running 32-bit executables and
    refuses on looking at them. I don't have any ARM Windows 10 left to try
    them on.


    I am in two minds about this.

    On the one hand, I strongly believe in backwards compatibility.

    On the other hand, we are only a little over 11 years to the physical
    reality that is 2038.

    BTW, I also worry that because we did such a good job with Y2K, because
    we got the resources and people we needed in plenty of time due to the publicity, that people may not take 2038 seriously enough until it is too
    late and the disaster is upon us.

    Simon.
    --
    Simon Clubley, clubley@remove_me.eisner.decus.org-Earth.UFP
    Walking destinations on a map are further away than they appear.
    --- Synchronet 3.22a-Linux NewsLink 1.2
  • From cross@cross@spitfire.i.gajendra.net (Dan Cross) to comp.os.vms on Mon Oct 5 14:16:30 2026
    From Newsgroup: comp.os.vms

    In article <11a04vp$tous$1@dont-email.me>,
    Simon Clubley <clubley@remove_me.eisner.decus.org-Earth.UFP> wrote:
    On 2026-10-04, John Dallman <jgd@cix.co.uk> wrote:
    In article <memo.20251007223453.10624Y@jgd.cix.co.uk>, jgd@cix.co.uk
    (John Dallman) wrote:

    but if Windows is only using A64 it doesn't matter.

    Microsoft supply compilers that can target 32-bit code, and run-time
    libraries for 32-bit programs. I've never tried building anything
    on ARM Windows for 32-bit so I don't know how well they work. I
    don't know if ARM Windows 11, which is always a 64-bit OS, will
    notice that the hardware is incapable of running A32/T32 ...

    ARM Windows 11 is not interested in running 32-bit executables and
    refuses on looking at them. I don't have any ARM Windows 10 left to try
    them on.


    I am in two minds about this.

    On the one hand, I strongly believe in backwards compatibility.

    On the other hand, we are only a little over 11 years to the physical
    reality that is 2038.

    Is time on 32-bit Windows measured as a signed 32-bit integer
    tracking seconds since Jan 1, 1970, UTC? Wikipedia suggests
    something like that.

    If correct, then yes, it is a problem. If not, then I don't see
    any reason why one could not use a 32-bit executable with 64-bit
    `time_t`.

    BTW, I also worry that because we did such a good job with Y2K, because
    we got the resources and people we needed in plenty of time due to the >publicity, that people may not take 2038 seriously enough until it is too >late and the disaster is upon us.

    What I most worry about is some random SCADA system in a closet
    or utility shed somewhere that no one has bothered to look at in
    15 years.

    - Dan C.

    --- Synchronet 3.22a-Linux NewsLink 1.2
  • From cross@cross@spitfire.i.gajendra.net (Dan Cross) to comp.os.vms on Mon Oct 5 14:50:15 2026
    From Newsgroup: comp.os.vms

    In article <11a0d4k$11ecp$1@dont-email.me>,
    Arne Vajh|+j <arne@vajhoej.dk> wrote:
    On 10/5/2026 10:16 AM, Dan Cross wrote:
    In article <11a04vp$tous$1@dont-email.me>,
    Simon Clubley <clubley@remove_me.eisner.decus.org-Earth.UFP> wrote:
    On the other hand, we are only a little over 11 years to the physical
    reality that is 2038.

    Is time on 32-bit Windows measured as a signed 32-bit integer
    tracking seconds since Jan 1, 1970, UTC? Wikipedia suggests
    something like that.

    If correct, then yes, it is a problem. If not, then I don't see
    any reason why one could not use a 32-bit executable with 64-bit
    `time_t`.
    Windows NT is more VMS than Unix. Internally it use 64 bit
    with 100 ns units since 1601 (not 1858).

    C time_t switched to 64 bit as default on both 32 bit and 64 bit
    Windows two decades ago.

    Of course someone can have very old binaries or newer
    binaries that for whatever reason has been built with
    _USE_32BIT_TIME_T.

    From https://en.wikipedia.org/wiki/Year_2038_problem#Vulnerable_systems,

    |As of circa Windows 10 version 22H2, if an x86 (32-bit) device
    |has not been turned on at the rollover point, but is started up
    |after the rollover, the system clock can skip ahead to 12 April
    |2160, and the date options in Windows 10's main settings menu
    |does not show any selectable numbers.[13]

    This suggests they're doing the same, "treat a 32-bit int as
    unsigned" trick that Kick McKusick and the FreeBSD people did
    for UFS2, but have a bug somewhere.

    - Dan C.

    --- Synchronet 3.22a-Linux NewsLink 1.2
  • From =?UTF-8?Q?Arne_Vajh=C3=B8j?=@arne@vajhoej.dk to comp.os.vms on Mon Oct 5 10:44:35 2026
    From Newsgroup: comp.os.vms

    On 10/5/2026 10:16 AM, Dan Cross wrote:
    In article <11a04vp$tous$1@dont-email.me>,
    Simon Clubley <clubley@remove_me.eisner.decus.org-Earth.UFP> wrote:
    On the other hand, we are only a little over 11 years to the physical
    reality that is 2038.

    Is time on 32-bit Windows measured as a signed 32-bit integer
    tracking seconds since Jan 1, 1970, UTC? Wikipedia suggests
    something like that.

    If correct, then yes, it is a problem. If not, then I don't see
    any reason why one could not use a 32-bit executable with 64-bit
    `time_t`.
    Windows NT is more VMS than Unix. Internally it use 64 bit
    with 100 ns units since 1601 (not 1858).

    C time_t switched to 64 bit as default on both 32 bit and 64 bit
    Windows two decades ago.

    Of course someone can have very old binaries or newer
    binaries that for whatever reason has been built with
    _USE_32BIT_TIME_T.

    Arne

    --- Synchronet 3.22a-Linux NewsLink 1.2
  • From =?UTF-8?Q?Arne_Vajh=C3=B8j?=@arne@vajhoej.dk to comp.os.vms on Mon Oct 5 10:52:28 2026
    From Newsgroup: comp.os.vms

    On 10/5/2026 10:44 AM, Arne Vajh|+j wrote:
    On 10/5/2026 10:16 AM, Dan Cross wrote:
    In article <11a04vp$tous$1@dont-email.me>,
    Simon Clubley-a <clubley@remove_me.eisner.decus.org-Earth.UFP> wrote:
    On the other hand, we are only a little over 11 years to the physical
    reality that is 2038.

    Is time on 32-bit Windows measured as a signed 32-bit integer
    tracking seconds since Jan 1, 1970, UTC?-a Wikipedia suggests
    something like that.

    If correct, then yes, it is a problem.-a If not, then I don't see
    any reason why one could not use a 32-bit executable with 64-bit
    `time_t`.
    Windows NT is more VMS than Unix. Internally it use 64 bit
    with 100 ns units since 1601 (not 1858).

    C time_t switched to 64 bit as default on both 32 bit and 64 bit
    Windows two decades ago.

    Of course someone can have very old binaries or newer
    binaries that for whatever reason has been built with
    _USE_32BIT_TIME_T.
    And in case someone is interested in VMS then it is:

    100 ns units since 1858 internally and 32 bit unsigned
    time_t.

    So VMS has a 2106 problem not a 2038 problem.

    Arne

    --- Synchronet 3.22a-Linux NewsLink 1.2
  • From kludge@kludge@panix.com (Scott Dorsey) to comp.os.vms on Mon Oct 5 12:31:47 2026
    From Newsgroup: comp.os.vms

    Dan Cross <cross@spitfire.i.gajendra.net> wrote:

    Is time on 32-bit Windows measured as a signed 32-bit integer
    tracking seconds since Jan 1, 1970, UTC? Wikipedia suggests
    something like that.

    No, Windows time is not epoch time. And Windows uses several different
    time formats for different things. SYSTEMTIME is stored as a weird and
    sloppy structure, but the times in the filesystem are stored in a totally different format.
    --scott
    --
    "C'est un Nagra. C'est suisse, et tres, tres precis."
    --- Synchronet 3.22a-Linux NewsLink 1.2
  • From =?UTF-8?Q?Arne_Vajh=C3=B8j?=@arne@vajhoej.dk to comp.os.vms on Mon Oct 5 12:51:33 2026
    From Newsgroup: comp.os.vms

    On 10/5/2026 12:31 PM, Scott Dorsey wrote:
    Dan Cross <cross@spitfire.i.gajendra.net> wrote:
    Is time on 32-bit Windows measured as a signed 32-bit integer
    tracking seconds since Jan 1, 1970, UTC? Wikipedia suggests
    something like that.

    No, Windows time is not epoch time. And Windows uses several different
    time formats for different things. SYSTEMTIME is stored as a weird and sloppy structure, but the times in the filesystem are stored in a totally different format.
    SYSTEMTIME is more like struct tm than like time_t.

    FILETIME is almost a plain 64 bit integer - it is
    just declared as two 32 bit for historic reasons.
    But will anyone here say that they have not called
    SYS$GETTIM with two 32 bit?? :-)

    Arne



    --- Synchronet 3.22a-Linux NewsLink 1.2
  • From Michael S@already5chosen@yahoo.com to comp.os.vms on Mon Oct 5 22:22:14 2026
    From Newsgroup: comp.os.vms

    On Mon, 5 Oct 2026 14:50:15 -0000 (UTC)
    cross@spitfire.i.gajendra.net (Dan Cross) wrote:
    In article <11a0d4k$11ecp$1@dont-email.me>,
    Arne Vajh|+j <arne@vajhoej.dk> wrote:
    On 10/5/2026 10:16 AM, Dan Cross wrote:
    In article <11a04vp$tous$1@dont-email.me>,
    Simon Clubley <clubley@remove_me.eisner.decus.org-Earth.UFP>
    wrote:
    On the other hand, we are only a little over 11 years to the
    physical reality that is 2038.

    Is time on 32-bit Windows measured as a signed 32-bit integer
    tracking seconds since Jan 1, 1970, UTC? Wikipedia suggests
    something like that.

    If correct, then yes, it is a problem. If not, then I don't see
    any reason why one could not use a 32-bit executable with 64-bit
    `time_t`.
    Windows NT is more VMS than Unix. Internally it use 64 bit
    with 100 ns units since 1601 (not 1858).

    C time_t switched to 64 bit as default on both 32 bit and 64 bit
    Windows two decades ago.

    Of course someone can have very old binaries or newer
    binaries that for whatever reason has been built with
    _USE_32BIT_TIME_T.

    From
    https://en.wikipedia.org/wiki/Year_2038_problem#Vulnerable_systems,

    |As of circa Windows 10 version 22H2, if an x86 (32-bit) device
    |has not been turned on at the rollover point, but is started up
    |after the rollover, the system clock can skip ahead to 12 April
    |2160, and the date options in Windows 10's main settings menu
    |does not show any selectable numbers.[13]

    This suggests they're doing the same, "treat a 32-bit int as
    unsigned" trick that Kick McKusick and the FreeBSD people did
    for UFS2, but have a bug somewhere.

    - Dan C.

    I did not read the article, but from citation it sounds that they are
    talking about default C RTL DLLs supplied with the OS, most likely msvcr120.dll, rather than about OS itself.
    As mentioned by few posters above, Windows native time format (FILETIME)
    is identical in 32-bit and 64-bit OS variants. It is going to overflow
    in 308th century AC. So it does not sound like a problem that even the
    youngest among us should be concerned about.
    --- Synchronet 3.22a-Linux NewsLink 1.2
  • From Michael S@already5chosen@yahoo.com to comp.os.vms on Mon Oct 5 22:47:17 2026
    From Newsgroup: comp.os.vms

    On Mon, 5 Oct 2026 10:52:28 -0400
    Arne Vajhoj <arne@vajhoej.dk> wrote:
    On 10/5/2026 10:44 AM, Arne Vajhoj wrote:
    And in case someone is interested in VMS then it is:

    100 ns units since 1858 internally and 32 bit unsigned
    time_t.

    So VMS has a 2106 problem not a 2038 problem.

    Arne

    I can not figure out what you have in mind.
    --- Synchronet 3.22a-Linux NewsLink 1.2
  • From jgd@jgd@cix.co.uk (John Dallman) to comp.os.vms on Mon Oct 5 22:10:40 2026
    From Newsgroup: comp.os.vms

    In article <11a0bfu$lkg$1@reader2.panix.com>,
    cross@spitfire.i.gajendra.net (Dan Cross) wrote:

    Is time on 32-bit Windows measured as a signed 32-bit integer
    tracking seconds since Jan 1, 1970, UTC? Wikipedia suggests
    something like that.

    If correct, then yes, it is a problem. If not, then I don't see
    any reason why one could not use a 32-bit executable with 64-bit
    `time_t`.

    As of Visual Studio 2022, and several previous versions, time_t is 64-bit
    on 32- and 64-bit Windows. The implementation in the run-time libraries
    may be defective, of course.

    John
    --- Synchronet 3.22a-Linux NewsLink 1.2
  • From cross@cross@spitfire.i.gajendra.net (Dan Cross) to comp.os.vms on Mon Oct 5 21:34:58 2026
    From Newsgroup: comp.os.vms

    In article <memo.20261005220922.6844b@jgd.cix.co.uk>,
    John Dallman <jgd@cix.co.uk> wrote:
    In article <11a0bfu$lkg$1@reader2.panix.com>,
    cross@spitfire.i.gajendra.net (Dan Cross) wrote:

    Is time on 32-bit Windows measured as a signed 32-bit integer
    tracking seconds since Jan 1, 1970, UTC? Wikipedia suggests
    something like that.

    If correct, then yes, it is a problem. If not, then I don't see
    any reason why one could not use a 32-bit executable with 64-bit
    `time_t`.

    As of Visual Studio 2022, and several previous versions, time_t is 64-bit
    on 32- and 64-bit Windows. The implementation in the run-time libraries
    may be defective, of course.

    The original question was about windows on ARM; I suspect that
    port happened after 2038 issues were well understood, in which
    case, I would expect the ABI to use a 64-bit `time_t` on 32-bit
    platforms (or make it unsigned). It would be odd to me that one
    would field a new platform in, what, the 2010s or something,
    using a 32-bit signed `time_t` susceptible to 2038 problems.

    Then again, I don't know that many people saw 64-bit ARM
    becoming as important as it has. If your use-case was a cell
    phone that was going to be obsolete in 5 years anyway, perhaps
    one is not as rigorous in defining forward-looking interface to
    ensure backwards compatibilty on a 25-year time scale.

    - Dan C.

    --- Synchronet 3.22a-Linux NewsLink 1.2
  • From cross@cross@spitfire.i.gajendra.net (Dan Cross) to comp.os.vms on Wed Oct 7 11:58:27 2026
    From Newsgroup: comp.os.vms

    In article <20261005224717.00004719@yahoo.com>,
    Michael S <already5chosen@yahoo.com> wrote:
    On Mon, 5 Oct 2026 10:52:28 -0400
    Arne Vajhoj <arne@vajhoej.dk> wrote:

    On 10/5/2026 10:44 AM, Arne Vajhoj wrote:
    And in case someone is interested in VMS then it is:

    100 ns units since 1858 internally and 32 bit unsigned
    time_t.

    So VMS has a 2106 problem not a 2038 problem.

    I can not figure out what you have in mind.

    He's referring to when VMS's `time_t` rolls over. Since it is
    an unsigned 32-bit int, overflow will be in 2106, not 2038, as
    in the case for signed 32-bit `time_t` on other systems.

    It is true that this is not something that the _operating
    system_ itself cares about, as it's in the C RTL (or anything
    that uses the Unix epoch), but it's a problem for VMS in the
    sense that it still needs to be fixed, tested, and shipped in
    less than 100 years.

    - Dan C.

    --- Synchronet 3.22a-Linux NewsLink 1.2
  • From Michael S@already5chosen@yahoo.com to comp.os.vms on Wed Oct 7 22:15:31 2026
    From Newsgroup: comp.os.vms

    On Wed, 7 Oct 2026 11:58:27 -0000 (UTC)
    cross@spitfire.i.gajendra.net (Dan Cross) wrote:
    In article <20261005224717.00004719@yahoo.com>,
    Michael S <already5chosen@yahoo.com> wrote:
    On Mon, 5 Oct 2026 10:52:28 -0400
    Arne Vajhoj <arne@vajhoej.dk> wrote:

    On 10/5/2026 10:44 AM, Arne Vajhoj wrote:
    And in case someone is interested in VMS then it is:

    100 ns units since 1858 internally and 32 bit unsigned
    time_t.

    So VMS has a 2106 problem not a 2038 problem.

    I can not figure out what you have in mind.

    He's referring to when VMS's `time_t` rolls over. Since it is
    an unsigned 32-bit int, overflow will be in 2106, not 2038, as
    in the case for signed 32-bit `time_t` on other systems.

    What I didn't get is how the problem related to 1858.
    Now I understand that it is not related.
    It is true that this is not something that the _operating
    system_ itself cares about, as it's in the C RTL (or anything
    that uses the Unix epoch), but it's a problem for VMS in the
    sense that it still needs to be fixed, tested, and shipped in
    less than 100 years.

    - Dan C.

    --- Synchronet 3.22a-Linux NewsLink 1.2
  • From scott@scott@slp53.sl.home (Scott Lurndal) to comp.os.vms on Wed Oct 7 20:54:18 2026
    From Newsgroup: comp.os.vms

    Michael S <already5chosen@yahoo.com> writes:
    On Wed, 7 Oct 2026 11:58:27 -0000 (UTC)
    cross@spitfire.i.gajendra.net (Dan Cross) wrote:

    In article <20261005224717.00004719@yahoo.com>,
    Michael S <already5chosen@yahoo.com> wrote:
    On Mon, 5 Oct 2026 10:52:28 -0400
    Arne Vajh=F8j <arne@vajhoej.dk> wrote:
    =20
    On 10/5/2026 10:44 AM, Arne Vajh=F8j wrote:
    And in case someone is interested in VMS then it is:
    =20
    100 ns units since 1858 internally and 32 bit unsigned
    time_t.
    =20
    So VMS has a 2106 problem not a 2038 problem. =20

    I can not figure out what you have in mind. =20
    =20
    He's referring to when VMS's `time_t` rolls over. Since it is
    an unsigned 32-bit int, overflow will be in 2106, not 2038, as
    in the case for signed 32-bit `time_t` on other systems.
    =20

    What I didn't get is how the problem related to 1858.
    Now I understand that it is not related.

    Peripherally related, perhaps. The baseline for VMS
    time is 17 November 1858 00:00:00.

    --- Synchronet 3.22a-Linux NewsLink 1.2
  • From Chris Townley@news@cct-net.co.uk to comp.os.vms on Wed Oct 7 22:44:10 2026
    From Newsgroup: comp.os.vms

    On 07/10/2026 21:54, Scott Lurndal wrote:
    Michael S <already5chosen@yahoo.com> writes:
    On Wed, 7 Oct 2026 11:58:27 -0000 (UTC)
    cross@spitfire.i.gajendra.net (Dan Cross) wrote:

    In article <20261005224717.00004719@yahoo.com>,
    Michael S <already5chosen@yahoo.com> wrote:
    On Mon, 5 Oct 2026 10:52:28 -0400
    Arne Vajh=F8j <arne@vajhoej.dk> wrote:
    =20
    On 10/5/2026 10:44 AM, Arne Vajh=F8j wrote:
    And in case someone is interested in VMS then it is:
    =20
    100 ns units since 1858 internally and 32 bit unsigned
    time_t.
    =20
    So VMS has a 2106 problem not a 2038 problem. =20

    I can not figure out what you have in mind. =20
    =20
    He's referring to when VMS's `time_t` rolls over. Since it is
    an unsigned 32-bit int, overflow will be in 2106, not 2038, as
    in the case for signed 32-bit `time_t` on other systems.
    =20

    What I didn't get is how the problem related to 1858.
    Now I understand that it is not related.

    Peripherally related, perhaps. The baseline for VMS
    time is 17 November 1858 00:00:00.

    For Vax basic users, it was 1970, so we had to fix it before 2003 - we
    did it with Y2K
    --
    Chris
    --- Synchronet 3.22a-Linux NewsLink 1.2
  • From scott@scott@slp53.sl.home (Scott Lurndal) to comp.os.vms on Wed Oct 7 22:14:23 2026
    From Newsgroup: comp.os.vms

    Chris Townley <news@cct-net.co.uk> writes:
    On 07/10/2026 21:54, Scott Lurndal wrote:
    Michael S <already5chosen@yahoo.com> writes:
    On Wed, 7 Oct 2026 11:58:27 -0000 (UTC)
    cross@spitfire.i.gajendra.net (Dan Cross) wrote:

    In article <20261005224717.00004719@yahoo.com>,
    Michael S <already5chosen@yahoo.com> wrote:
    On Mon, 5 Oct 2026 10:52:28 -0400
    Arne Vajh=F8j <arne@vajhoej.dk> wrote:
    =20
    On 10/5/2026 10:44 AM, Arne Vajh=F8j wrote:
    And in case someone is interested in VMS then it is:
    =20
    100 ns units since 1858 internally and 32 bit unsigned
    time_t.
    =20
    So VMS has a 2106 problem not a 2038 problem. =20

    I can not figure out what you have in mind. =20
    =20
    He's referring to when VMS's `time_t` rolls over. Since it is
    an unsigned 32-bit int, overflow will be in 2106, not 2038, as
    in the case for signed 32-bit `time_t` on other systems.
    =20

    What I didn't get is how the problem related to 1858.
    Now I understand that it is not related.

    Peripherally related, perhaps. The baseline for VMS
    time is 17 November 1858 00:00:00.

    For Vax basic users, it was 1970, so we had to fix it before 2003 - we
    did it with Y2K

    That seems strange, unless you were running your BASIC
    interpreter on UnixV32 or BSD (the unix epoch started
    Dec 31, 1969 at midnight).

    TIME function in VSI BASIC seems to have been 'number of seconds since midnight' as
    a floating point value. I mainly used Macro32, Bliss32 and VAX Pascal, where I had access
    to the VMS starlet functions for time management and inquiry.

    VSI BASIC Returns

    0 The amount of time elapsed since midnight in seconds
    1 The CPU time of the current process in tenths of a second
    2 The connect time of the current process in minutes

    TIME$ returns a string with the current time.
    --- Synchronet 3.22a-Linux NewsLink 1.2
  • From Chris Townley@news@cct-net.co.uk to comp.os.vms on Wed Oct 7 23:53:47 2026
    From Newsgroup: comp.os.vms

    On 07/10/2026 23:14, Scott Lurndal wrote:
    Chris Townley <news@cct-net.co.uk> writes:
    On 07/10/2026 21:54, Scott Lurndal wrote:
    Michael S <already5chosen@yahoo.com> writes:
    On Wed, 7 Oct 2026 11:58:27 -0000 (UTC)
    cross@spitfire.i.gajendra.net (Dan Cross) wrote:

    In article <20261005224717.00004719@yahoo.com>,
    Michael S <already5chosen@yahoo.com> wrote:
    On Mon, 5 Oct 2026 10:52:28 -0400
    Arne Vajh=F8j <arne@vajhoej.dk> wrote:
    =20
    On 10/5/2026 10:44 AM, Arne Vajh=F8j wrote:
    And in case someone is interested in VMS then it is:
    =20
    100 ns units since 1858 internally and 32 bit unsigned
    time_t.
    =20
    So VMS has a 2106 problem not a 2038 problem. =20

    I can not figure out what you have in mind. =20
    =20
    He's referring to when VMS's `time_t` rolls over. Since it is
    an unsigned 32-bit int, overflow will be in 2106, not 2038, as
    in the case for signed 32-bit `time_t` on other systems.
    =20

    What I didn't get is how the problem related to 1858.
    Now I understand that it is not related.

    Peripherally related, perhaps. The baseline for VMS
    time is 17 November 1858 00:00:00.

    For Vax basic users, it was 1970, so we had to fix it before 2003 - we
    did it with Y2K

    That seems strange, unless you were running your BASIC
    interpreter on UnixV32 or BSD (the unix epoch started
    Dec 31, 1969 at midnight).

    TIME function in VSI BASIC seems to have been 'number of seconds since midnight' as
    a floating point value. I mainly used Macro32, Bliss32 and VAX Pascal, where I had access
    to the VMS starlet functions for time management and inquiry.

    VSI BASIC Returns

    0 The amount of time elapsed since midnight in seconds
    1 The CPU time of the current process in tenths of a second
    2 The connect time of the current process in minutes

    TIME$ returns a string with the current time.

    Our dates were stored as Julian dates
    --
    Chris
    --- Synchronet 3.22a-Linux NewsLink 1.2
  • From scott@scott@slp53.sl.home (Scott Lurndal) to comp.os.vms on Thu Oct 8 15:50:26 2026
    From Newsgroup: comp.os.vms

    Chris Townley <news@cct-net.co.uk> writes:
    On 07/10/2026 23:14, Scott Lurndal wrote:
    Chris Townley <news@cct-net.co.uk> writes:
    On 07/10/2026 21:54, Scott Lurndal wrote:
    Michael S <already5chosen@yahoo.com> writes:
    On Wed, 7 Oct 2026 11:58:27 -0000 (UTC)
    cross@spitfire.i.gajendra.net (Dan Cross) wrote:

    In article <20261005224717.00004719@yahoo.com>,
    Michael S <already5chosen@yahoo.com> wrote:
    On Mon, 5 Oct 2026 10:52:28 -0400
    Arne Vajh=F8j <arne@vajhoej.dk> wrote:
    =20
    On 10/5/2026 10:44 AM, Arne Vajh=F8j wrote:
    And in case someone is interested in VMS then it is:
    =20
    100 ns units since 1858 internally and 32 bit unsigned
    time_t.
    =20
    So VMS has a 2106 problem not a 2038 problem. =20

    I can not figure out what you have in mind. =20
    =20
    He's referring to when VMS's `time_t` rolls over. Since it is
    an unsigned 32-bit int, overflow will be in 2106, not 2038, as
    in the case for signed 32-bit `time_t` on other systems.
    =20

    What I didn't get is how the problem related to 1858.
    Now I understand that it is not related.

    Peripherally related, perhaps. The baseline for VMS
    time is 17 November 1858 00:00:00.

    For Vax basic users, it was 1970, so we had to fix it before 2003 - we
    did it with Y2K

    That seems strange, unless you were running your BASIC
    interpreter on UnixV32 or BSD (the unix epoch started
    Dec 31, 1969 at midnight).

    TIME function in VSI BASIC seems to have been 'number of seconds since midnight' as
    a floating point value. I mainly used Macro32, Bliss32 and VAX Pascal, where I had access
    to the VMS starlet functions for time management and inquiry.

    VSI BASIC Returns

    0 The amount of time elapsed since midnight in seconds
    1 The CPU time of the current process in tenths of a second
    2 The connect time of the current process in minutes

    TIME$ returns a string with the current time.

    Our dates were stored as Julian dates

    Not uncommon in the mainframe days. Did you use 5 or 7 digit
    values? (YYDDD or YYYYDDD)
    --- Synchronet 3.22a-Linux NewsLink 1.2
  • From Chris Townley@news@cct-net.co.uk to comp.os.vms on Thu Oct 8 17:26:15 2026
    From Newsgroup: comp.os.vms

    On 08/10/2026 16:50, Scott Lurndal wrote:
    Chris Townley <news@cct-net.co.uk> writes:
    On 07/10/2026 23:14, Scott Lurndal wrote:
    Chris Townley <news@cct-net.co.uk> writes:
    On 07/10/2026 21:54, Scott Lurndal wrote:
    Michael S <already5chosen@yahoo.com> writes:
    On Wed, 7 Oct 2026 11:58:27 -0000 (UTC)
    cross@spitfire.i.gajendra.net (Dan Cross) wrote:

    In article <20261005224717.00004719@yahoo.com>,
    Michael S <already5chosen@yahoo.com> wrote:
    On Mon, 5 Oct 2026 10:52:28 -0400
    Arne Vajh=F8j <arne@vajhoej.dk> wrote:
    =20
    On 10/5/2026 10:44 AM, Arne Vajh=F8j wrote:
    And in case someone is interested in VMS then it is:
    =20
    100 ns units since 1858 internally and 32 bit unsigned
    time_t.
    =20
    So VMS has a 2106 problem not a 2038 problem. =20

    I can not figure out what you have in mind. =20
    =20
    He's referring to when VMS's `time_t` rolls over. Since it is
    an unsigned 32-bit int, overflow will be in 2106, not 2038, as
    in the case for signed 32-bit `time_t` on other systems.
    =20

    What I didn't get is how the problem related to 1858.
    Now I understand that it is not related.

    Peripherally related, perhaps. The baseline for VMS
    time is 17 November 1858 00:00:00.

    For Vax basic users, it was 1970, so we had to fix it before 2003 - we >>>> did it with Y2K

    That seems strange, unless you were running your BASIC
    interpreter on UnixV32 or BSD (the unix epoch started
    Dec 31, 1969 at midnight).

    TIME function in VSI BASIC seems to have been 'number of seconds since midnight' as
    a floating point value. I mainly used Macro32, Bliss32 and VAX Pascal, where I had access
    to the VMS starlet functions for time management and inquiry.

    VSI BASIC Returns

    0 The amount of time elapsed since midnight in seconds
    1 The CPU time of the current process in tenths of a second
    2 The connect time of the current process in minutes

    TIME$ returns a string with the current time.

    Our dates were stored as Julian dates

    Not uncommon in the mainframe days. Did you use 5 or 7 digit
    values? (YYDDD or YYYYDDD)

    Dates were Julian dates, Year starting 01-Jan-1970 X 1000, plus days
    and stored as signed word - Basic doesn't do unsigned integers
    --
    Chris
    --- Synchronet 3.22a-Linux NewsLink 1.2
  • From Rich Alderson@news@alderson.users.panix.com to comp.os.vms on Thu Oct 8 17:14:22 2026
    From Newsgroup: comp.os.vms

    Chris Townley <news@cct-net.co.uk> writes:

    On 08/10/2026 16:50, Scott Lurndal wrote:
    Chris Townley <news@cct-net.co.uk> writes:
    On 07/10/2026 23:14, Scott Lurndal wrote:
    Chris Townley <news@cct-net.co.uk> writes:

    [ massive snippage ]

    Our dates were stored as Julian dates

    Not uncommon in the mainframe days. Did you use 5 or 7 digit values?
    (YYDDD or YYYYDDD)

    Dates were Julian dates, Year starting 01-Jan-1970 X 1000, plus days
    and stored as signed word - Basic doesn't do unsigned integers

    That's not "Julian dates". Julian dates are an integer number of days since the beginning of the Julian epoch (12:00 noon, 1 January 4713 Julian =
    24 November 4714 Gregorian) + the fraction of a day since the beginning of the Julian day.

    The TENEX > TOPS-20 > later VMS base of 17 November 1858 at midnight is called the "modified Julian date", being Julian date 2400000.5. That made for a safe epoch earlier than any automated data processing existed, but reasonable for keeping the day number in the left half of a 36 bit word, and the fraction in the right half.

    Julian dates start at noon because they were originally intended for astronomical records, such that breaking between two calendar dates for a single night's observations was an annoyance at the very least. The modification to a midnight start for business data processing is an obvious change.
    --
    Rich Alderson news@alderson.users.panix.com
    Audendum est, et veritas investiganda; quam etiamsi non assequamur,
    omnino tamen proprius, quam nunc sumus, ad eam perveniemus.
    --Galen --- Synchronet 3.22a-Linux NewsLink 1.2
  • From scott@scott@slp53.sl.home (Scott Lurndal) to comp.os.vms on Fri Oct 9 16:44:58 2026
    From Newsgroup: comp.os.vms

    Rich Alderson <news@alderson.users.panix.com> writes:
    Chris Townley <news@cct-net.co.uk> writes:

    On 08/10/2026 16:50, Scott Lurndal wrote:
    Chris Townley <news@cct-net.co.uk> writes:
    On 07/10/2026 23:14, Scott Lurndal wrote:
    Chris Townley <news@cct-net.co.uk> writes:

    [ massive snippage ]

    Our dates were stored as Julian dates

    Not uncommon in the mainframe days. Did you use 5 or 7 digit values?
    (YYDDD or YYYYDDD)

    Dates were Julian dates, Year starting 01-Jan-1970 X 1000, plus days
    and stored as signed word - Basic doesn't do unsigned integers

    That's not "Julian dates". Julian dates are an integer number of days since >the beginning of the Julian epoch (12:00 noon, 1 January 4713 Julian =
    24 November 4714 Gregorian) + the fraction of a day since the beginning of the >Julian day.


    Technically speaking, perhaps. In reality, such notations were commonly _called_ Julian dates in computing during the 1960s through Y2K.

    --- Synchronet 3.22a-Linux NewsLink 1.2
  • From Rich Alderson@news@alderson.users.panix.com to comp.os.vms on Fri Oct 9 19:03:46 2026
    From Newsgroup: comp.os.vms

    scott@slp53.sl.home (Scott Lurndal) writes:

    Rich Alderson <news@alderson.users.panix.com> writes:

    [ snip ]

    That's not "Julian dates". Julian dates are an integer number of days since >> the beginning of the Julian epoch (12:00 noon, 1 January 4713 Julian =
    24 November 4714 Gregorian) + the fraction of a day since the beginning of >> the Julian day.

    Technically speaking, perhaps. In reality, such notations were commonly _called_ Julian dates in computing during the 1960s through Y2K.

    I know, Scott, but I got called out on that usage in the 20th Century. Thought I'd share the pain.
    --
    Rich Alderson news@alderson.users.panix.com
    Audendum est, et veritas investiganda; quam etiamsi non assequamur,
    omnino tamen proprius, quam nunc sumus, ad eam perveniemus.
    --Galen --- Synchronet 3.22a-Linux NewsLink 1.2
  • From Dave Froble@davef@tsoft-inc.com to comp.os.vms on Sat Oct 10 10:52:41 2026
    From Newsgroup: comp.os.vms

    On 10/7/2026 4:54 PM, Scott Lurndal wrote:
    Michael S <already5chosen@yahoo.com> writes:
    On Wed, 7 Oct 2026 11:58:27 -0000 (UTC)
    cross@spitfire.i.gajendra.net (Dan Cross) wrote:

    In article <20261005224717.00004719@yahoo.com>,
    Michael S <already5chosen@yahoo.com> wrote:
    On Mon, 5 Oct 2026 10:52:28 -0400
    Arne Vajh=F8j <arne@vajhoej.dk> wrote:
    =20
    On 10/5/2026 10:44 AM, Arne Vajh=F8j wrote:
    And in case someone is interested in VMS then it is:
    =20
    100 ns units since 1858 internally and 32 bit unsigned
    time_t.
    =20
    So VMS has a 2106 problem not a 2038 problem. =20

    I can not figure out what you have in mind. =20
    =20
    He's referring to when VMS's `time_t` rolls over. Since it is
    an unsigned 32-bit int, overflow will be in 2106, not 2038, as
    in the case for signed 32-bit `time_t` on other systems.
    =20

    What I didn't get is how the problem related to 1858.
    Now I understand that it is not related.

    Peripherally related, perhaps. The baseline for VMS
    time is 17 November 1858 00:00:00.


    In what time zone?

    :-)
    --
    David Froble Tel: 724-529-0450
    Dave Froble Enterprises, Inc. E-Mail: davef@tsoft-inc.com
    DFE Ultralights, Inc.
    170 Grimplin Road
    Vanderbilt, PA 15486
    --- Synchronet 3.22a-Linux NewsLink 1.2
  • From Simon Clubley@clubley@remove_me.eisner.decus.org-Earth.UFP to comp.os.vms on Thu Oct 15 18:12:14 2026
    From Newsgroup: comp.os.vms

    On 2026-10-05, Michael S <already5chosen@yahoo.com> wrote:

    I did not read the article, but from citation it sounds that they are
    talking about default C RTL DLLs supplied with the OS, most likely msvcr120.dll, rather than about OS itself.
    As mentioned by few posters above, Windows native time format (FILETIME)
    is identical in 32-bit and 64-bit OS variants. It is going to overflow
    in 308th century AC. So it does not sound like a problem that even the youngest among us should be concerned about.


    So while the issue on Windows is not as simple as I thought it was
    (and thanks for the references everyone), that still leaves any
    legacy 32-bit Windows applications themselves to be considered.

    How many Windows applications are still going to fail in Jan-2038 ?

    It's obviously the legacy embedded style systems that are going to get
    hit hardest. Windows is a part of that in, for example, the interactive
    turnkey world, but you also have applications running on other operating systems (headless or otherwise) or bare metal systems in a box somewhere.

    This, BTW, is one area where virtualising hardware (to protect old
    systems against hardware failure) does not get you out of trouble.

    Simon.

    PS: Here's a charming thought: in the case of applications, and depending
    on the nature of the failure mode, it's not when the date reaches Jan-2038
    that the problem begins. It's when there's a date of Jan-2038 or later that needs to be recorded or otherwise processed that the problems start. That
    means you are likely to start seeing failures several years before Jan-2038 itself is reached.

    On the one hand, if I am right, that gives us less time than we thought
    we had. On the other hand, in the final few years, it would give us an increasingly public alarm clock that we could use to hit others over
    the head with before the main event.
    --
    Simon Clubley, clubley@remove_me.eisner.decus.org-Earth.UFP
    Walking destinations on a map are further away than they appear.
    --- Synchronet 3.22a-Linux NewsLink 1.2
  • From Simon Clubley@clubley@remove_me.eisner.decus.org-Earth.UFP to comp.os.vms on Thu Oct 15 18:27:11 2026
    From Newsgroup: comp.os.vms

    On 2026-10-10, Dave Froble <davef@tsoft-inc.com> wrote:
    On 10/7/2026 4:54 PM, Scott Lurndal wrote:
    Michael S <already5chosen@yahoo.com> writes:

    Peripherally related, perhaps. The baseline for VMS
    time is 17 November 1858 00:00:00.


    In what time zone?

    :-)


    Historically GMT, now UTC.

    https://en.wikipedia.org/wiki/Julian_day

    Simon.
    --
    Simon Clubley, clubley@remove_me.eisner.decus.org-Earth.UFP
    Walking destinations on a map are further away than they appear.
    --- Synchronet 3.22a-Linux NewsLink 1.2
  • From Fred. Zwarts@F.Zwarts@HetNet.nl to comp.os.vms on Fri Oct 16 09:52:52 2026
    From Newsgroup: comp.os.vms

    Op 10.okt.2026 om 16:52 schreef Dave Froble:
    On 10/7/2026 4:54 PM, Scott Lurndal wrote:
    Michael S <already5chosen@yahoo.com> writes:
    On Wed, 7 Oct 2026 11:58:27 -0000 (UTC)
    cross@spitfire.i.gajendra.net (Dan Cross) wrote:

    In article <20261005224717.00004719@yahoo.com>,
    Michael S-a <already5chosen@yahoo.com> wrote:
    On Mon, 5 Oct 2026 10:52:28 -0400
    Arne Vajh=F8j <arne@vajhoej.dk> wrote:
    =20
    On 10/5/2026 10:44 AM, Arne Vajh=F8j wrote:
    And in case someone is interested in VMS then it is:
    =20
    100 ns units since 1858 internally and 32 bit unsigned
    time_t.
    =20
    So VMS has a 2106 problem not a 2038 problem. =20

    I can not figure out what you have in mind. =20
    =20
    He's referring to when VMS's `time_t` rolls over.-a Since it is
    an unsigned 32-bit int, overflow will be in 2106, not 2038, as
    in the case for signed 32-bit `time_t` on other systems.
    =20

    What I didn't get is how the problem related to 1858.
    Now I understand that it is not related.

    Peripherally related, perhaps.-a The baseline for VMS
    time is 17 November 1858 00:00:00.


    In what time zone?

    :-)


    The local time zone. Initially VMS did not know about time zones and
    most sites ran the clock in the local time zone.
    --- Synchronet 3.22a-Linux NewsLink 1.2
  • From Rich Alderson@news@alderson.users.panix.com to comp.os.vms on Fri Oct 16 15:32:51 2026
    From Newsgroup: comp.os.vms

    "Fred. Zwarts" <F.Zwarts@HetNet.nl> writes:

    Op 10.okt.2026 om 16:52 schreef Dave Froble:
    On 10/7/2026 4:54 PM, Scott Lurndal wrote:
    Michael S <already5chosen@yahoo.com> writes:
    On Wed, 7 Oct 2026 11:58:27 -0000 (UTC)
    cross@spitfire.i.gajendra.net (Dan Cross) wrote:

    In article <20261005224717.00004719@yahoo.com>,
    Michael S-a <already5chosen@yahoo.com> wrote:
    On Mon, 5 Oct 2026 10:52:28 -0400
    Arne Vajh=F8j <arne@vajhoej.dk> wrote:

    On 10/5/2026 10:44 AM, Arne Vajh=F8j wrote:
    And in case someone is interested in VMS then it is:

    100 ns units since 1858 internally and 32 bit unsigned
    time_t.

    So VMS has a 2106 problem not a 2038 problem. =20

    I can not figure out what you have in mind. =20

    He's referring to when VMS's `time_t` rolls over.-a Since it is
    an unsigned 32-bit int, overflow will be in 2106, not 2038, as
    in the case for signed 32-bit `time_t` on other systems.

    What I didn't get is how the problem related to 1858.
    Now I understand that it is not related.

    Peripherally related, perhaps.-a The baseline for VMS
    time is 17 November 1858 00:00:00.

    In what time zone?

    :-)

    The local time zone. Initially VMS did not know about time zones and
    most sites ran the clock in the local time zone.

    That's a failing of VMS systems management. In TOPS-20 and TENEX, timezones were built into the monitor, so that NOT setting the local timezone was an error. I can't imagine that that was not true of VMS, but I've been surprised there before.
    --
    Rich Alderson news@alderson.users.panix.com
    Audendum est, et veritas investiganda; quam etiamsi non assequamur,
    omnino tamen proprius, quam nunc sumus, ad eam perveniemus.
    --Galen --- Synchronet 3.22a-Linux NewsLink 1.2
  • From scott@scott@slp53.sl.home (Scott Lurndal) to comp.os.vms on Fri Oct 16 20:58:46 2026
    From Newsgroup: comp.os.vms

    Rich Alderson <news@alderson.users.panix.com> writes:
    "Fred. Zwarts" <F.Zwarts@HetNet.nl> writes:

    Op 10.okt.2026 om 16:52 schreef Dave Froble:
    On 10/7/2026 4:54 PM, Scott Lurndal wrote:
    Michael S <already5chosen@yahoo.com> writes:
    On Wed, 7 Oct 2026 11:58:27 -0000 (UTC)
    cross@spitfire.i.gajendra.net (Dan Cross) wrote:

    In article <20261005224717.00004719@yahoo.com>,
    Michael S-a <already5chosen@yahoo.com> wrote:
    On Mon, 5 Oct 2026 10:52:28 -0400
    Arne Vajh=F8j <arne@vajhoej.dk> wrote:

    On 10/5/2026 10:44 AM, Arne Vajh=F8j wrote:
    And in case someone is interested in VMS then it is:

    100 ns units since 1858 internally and 32 bit unsigned
    time_t.

    So VMS has a 2106 problem not a 2038 problem. =20

    I can not figure out what you have in mind. =20

    He's referring to when VMS's `time_t` rolls over.-a Since it is
    an unsigned 32-bit int, overflow will be in 2106, not 2038, as
    in the case for signed 32-bit `time_t` on other systems.

    What I didn't get is how the problem related to 1858.
    Now I understand that it is not related.

    Peripherally related, perhaps.-a The baseline for VMS
    time is 17 November 1858 00:00:00.

    In what time zone?

    :-)

    The local time zone. Initially VMS did not know about time zones and
    most sites ran the clock in the local time zone.

    That's a failing of VMS systems management. In TOPS-20 and TENEX, timezones >were built into the monitor, so that NOT setting the local timezone was an >error. I can't imagine that that was not true of VMS, but I've been surprised >there before.

    VMS, before 6.0 didn't have timezones, nor did it handle standard/daylight
    time changes - the system manager need to change the time during a maintenance window.


    --- Synchronet 3.22a-Linux NewsLink 1.2
  • From Lawrence =?iso-8859-13?q?D=FFOliveiro?=@ldo@nz.invalid to comp.os.vms on Fri Oct 16 21:58:28 2026
    From Newsgroup: comp.os.vms

    On Fri, 16 Oct 2026 09:52:52 +0200, Fred. Zwarts wrote:

    Initially VMS did not know about time zones and most sites ran the
    clock in the local time zone.

    I think that was true of every computer system back then -- except
    one: Unix.

    Was it really such a radical idea to keep system time internally in
    (what was then GMT, later became) UTC? Seems it was, since other
    systems were in no hurry to copy the concept. Even Microsoft Windows
    remains prone to off-by-one-hour bugs being triggered around
    daylight-saving shifts, to this day.
    --- Synchronet 3.22a-Linux NewsLink 1.2
  • From kludge@kludge@panix.com (Scott Dorsey) to comp.os.vms on Sat Oct 17 07:33:05 2026
    From Newsgroup: comp.os.vms

    Lawrence =?iso-8859-13?q?D=FFOliveiro?= <ldo@nz.invalid> wrote:
    Was it really such a radical idea to keep system time internally in
    (what was then GMT, later became) UTC? Seems it was, since other
    systems were in no hurry to copy the concept. Even Microsoft Windows
    remains prone to off-by-one-hour bugs being triggered around
    daylight-saving shifts, to this day.

    It was, and it came about at about the same time that big networks
    were appearing and people in different timezones were all using the
    same computer. I think a lot of people invented the concept
    simultaneously because it suddenly became useful.
    --scott
    --
    "C'est un Nagra. C'est suisse, et tres, tres precis."
    --- Synchronet 3.22a-Linux NewsLink 1.2
  • From Stephen Hoffman@seaohveh@hoffmanlabs.invalid to comp.os.vms on Sun Oct 18 19:05:02 2026
    From Newsgroup: comp.os.vms

    On 2026-10-05 14:52:28 +0000, Arne Vajhoj said:

    And in case someone is interested in VMS then it is:

    100 ns units since 1858 internally and 32 bit unsigned time_t.

    So VMS has a 2106 problem not a 2038 problem.

    time_t is variously signed, and it overflows.

    OpenVMS fixed at least one 2038 bug, incidentally identified during the
    Y2K review.

    However... IIRC 2038 was explicitly kept outside the DEC Y2K review and testing range.

    Whether there are other 2038 bugs lurking? Might want to check your
    servers, etc.

    Having been deep into the DEC-then-Compaq Y2K stuff, I'm not so sure
    about the certainness of this web page: https://www.zx.net.nz/mirror/h71000.www7.hp.com/2038.html
    --
    Pure Personal Opinion | HoffmanLabs LLC

    --- Synchronet 3.22a-Linux NewsLink 1.2