Discussion:
[Pdl-porters] update PDL perl version to 5.10.x
Chris Marshall
2013-11-20 11:42:24 UTC
Permalink
I propose moving to perl 5.10.x as the officially
supported perl version for general PDL development,
effective immediately.

I don't expect this to affect the legacy PDL users
as they often are using PDL versions back as far
as 2.4.3.

Comment, discussion, votes?
Chris
Craig DeForest
2013-11-20 14:19:03 UTC
Permalink
Why not bump to 5.12?

(mobile)


On Nov 20, 2013, at 4:42 AM, Chris Marshall <***@gmail.com> wrote:

> I propose moving to perl 5.10.x as the officially
> supported perl version for general PDL development,
> effective immediately.
>
> I don't expect this to affect the legacy PDL users
> as they often are using PDL versions back as far
> as 2.4.3.
>
> Comment, discussion, votes?
> Chris
>
> _______________________________________________
> PDL-porters mailing list
> PDL-***@jach.hawaii.edu
> http://mailman.jach.hawaii.edu/mailman/listinfo/pdl-porters
>
Chris Marshall
2013-11-21 00:11:50 UTC
Permalink
Without detailed information on who would be affected
by a change in the required/supported version of perl,
I would prefer to minimize disruption for PDL users.
For example, I do know that the older cygwin releases
used perl 5.10.x and a jump to 5.12 could make them
not able to use PDL.

That said, if there is a specific need that could be
addressed by jumping to 5.12.x, that could justify
the change. Anything in mind---I haven't seen anything
myself.

All is not bad, if we go to 5.10 support, we can finally
use 'say' ... :-)

--Chris


On Wed, Nov 20, 2013 at 9:19 AM, Craig DeForest
<***@boulder.swri.edu> wrote:
> Why not bump to 5.12?
>
> (mobile)
>
>
> On Nov 20, 2013, at 4:42 AM, Chris Marshall <***@gmail.com> wrote:
>
>> I propose moving to perl 5.10.x as the officially
>> supported perl version for general PDL development,
>> effective immediately.
>>
>> I don't expect this to affect the legacy PDL users
>> as they often are using PDL versions back as far
>> as 2.4.3.
>>
>> Comment, discussion, votes?
>> Chris
>>
>> _______________________________________________
>> PDL-porters mailing list
>> PDL-***@jach.hawaii.edu
>> http://mailman.jach.hawaii.edu/mailman/listinfo/pdl-porters
>>
David Mertens
2013-11-21 14:10:53 UTC
Permalink
I don't see a need to move to 5.12. I think I'll eventually advocate for
that because of the lexical keyword API, but I don't have anything for that
yet and I think that Zefram typically writes CPAN modules which provide the
same C interfaces for older Perls.

The biggest Perl feature I would like to see is user-level pragmatic
modules<http://perldoc.perl.org/perlpragma.html>,
which come with 5.10. And actually, what I really want is warmings from a
module<http://perldoc.perl.org/perllexwarn.html#Reporting-Warnings-from-a-Module>.
AFAICT, that's been available since 5.6, so I suppose I/we could have
started on that a long time ago. Heh.

David


On Wed, Nov 20, 2013 at 7:11 PM, Chris Marshall <***@gmail.com>wrote:

> Without detailed information on who would be affected
> by a change in the required/supported version of perl,
> I would prefer to minimize disruption for PDL users.
> For example, I do know that the older cygwin releases
> used perl 5.10.x and a jump to 5.12 could make them
> not able to use PDL.
>
> That said, if there is a specific need that could be
> addressed by jumping to 5.12.x, that could justify
> the change. Anything in mind---I haven't seen anything
> myself.
>
> All is not bad, if we go to 5.10 support, we can finally
> use 'say' ... :-)
>
> --Chris
>
>
> On Wed, Nov 20, 2013 at 9:19 AM, Craig DeForest
> <***@boulder.swri.edu> wrote:
> > Why not bump to 5.12?
> >
> > (mobile)
> >
> >
> > On Nov 20, 2013, at 4:42 AM, Chris Marshall <***@gmail.com>
> wrote:
> >
> >> I propose moving to perl 5.10.x as the officially
> >> supported perl version for general PDL development,
> >> effective immediately.
> >>
> >> I don't expect this to affect the legacy PDL users
> >> as they often are using PDL versions back as far
> >> as 2.4.3.
> >>
> >> Comment, discussion, votes?
> >> Chris
> >>
> >> _______________________________________________
> >> PDL-porters mailing list
> >> PDL-***@jach.hawaii.edu
> >> http://mailman.jach.hawaii.edu/mailman/listinfo/pdl-porters
> >>
>
> _______________________________________________
> Perldl mailing list
> ***@jach.hawaii.edu
> http://mailman.jach.hawaii.edu/mailman/listinfo/perldl
>



--
"Debugging is twice as hard as writing the code in the first place.
Therefore, if you write the code as cleverly as possible, you are,
by definition, not smart enough to debug it." -- Brian Kernighan
David Mertens
2013-11-21 14:20:00 UTC
Permalink
That said, I think it might be nice to provide a clear dependency policy.
For example, we could promise to support all default Perls on the latest
versions of Cygwin, Strawberry, Active Perl, Mac OSX, Debian stable,
Centos, and Fedora. Note that some of those---namely Centos and Debian
stable---may be a bit conservative and we could instead provide
instructions or even install scripts that would install perlbrew and a
newer version of Perl.

Or, we could just promise to work on the *five* latest stable versions of
Perl (which, at the moment, would be 5.18, 5.16, 5.14, 5.12, and 5.10), and
therefore (try to) nudge Cygwin to move along. If we keep working with the
oldest Perl, then the distributors have no need to move forward, right?

David


On Thu, Nov 21, 2013 at 9:10 AM, David Mertens <***@gmail.com>wrote:

> I don't see a need to move to 5.12. I think I'll eventually advocate for
> that because of the lexical keyword API, but I don't have anything for that
> yet and I think that Zefram typically writes CPAN modules which provide the
> same C interfaces for older Perls.
>
> The biggest Perl feature I would like to see is user-level pragmatic
> modules <http://perldoc.perl.org/perlpragma.html>, which come with 5.10.
> And actually, what I really want is warmings from a module<http://perldoc.perl.org/perllexwarn.html#Reporting-Warnings-from-a-Module>.
> AFAICT, that's been available since 5.6, so I suppose I/we could have
> started on that a long time ago. Heh.
>
> David
>
>
> On Wed, Nov 20, 2013 at 7:11 PM, Chris Marshall <***@gmail.com>wrote:
>
>> Without detailed information on who would be affected
>> by a change in the required/supported version of perl,
>> I would prefer to minimize disruption for PDL users.
>> For example, I do know that the older cygwin releases
>> used perl 5.10.x and a jump to 5.12 could make them
>> not able to use PDL.
>>
>> That said, if there is a specific need that could be
>> addressed by jumping to 5.12.x, that could justify
>> the change. Anything in mind---I haven't seen anything
>> myself.
>>
>> All is not bad, if we go to 5.10 support, we can finally
>> use 'say' ... :-)
>>
>> --Chris
>>
>>
>> On Wed, Nov 20, 2013 at 9:19 AM, Craig DeForest
>> <***@boulder.swri.edu> wrote:
>> > Why not bump to 5.12?
>> >
>> > (mobile)
>> >
>> >
>> > On Nov 20, 2013, at 4:42 AM, Chris Marshall <***@gmail.com>
>> wrote:
>> >
>> >> I propose moving to perl 5.10.x as the officially
>> >> supported perl version for general PDL development,
>> >> effective immediately.
>> >>
>> >> I don't expect this to affect the legacy PDL users
>> >> as they often are using PDL versions back as far
>> >> as 2.4.3.
>> >>
>> >> Comment, discussion, votes?
>> >> Chris
>> >>
>> >> _______________________________________________
>> >> PDL-porters mailing list
>> >> PDL-***@jach.hawaii.edu
>> >> http://mailman.jach.hawaii.edu/mailman/listinfo/pdl-porters
>> >>
>>
>> _______________________________________________
>> Perldl mailing list
>> ***@jach.hawaii.edu
>> http://mailman.jach.hawaii.edu/mailman/listinfo/perldl
>>
>
>
>
> --
> "Debugging is twice as hard as writing the code in the first place.
> Therefore, if you write the code as cleverly as possible, you are,
> by definition, not smart enough to debug it." -- Brian Kernighan
>



--
"Debugging is twice as hard as writing the code in the first place.
Therefore, if you write the code as cleverly as possible, you are,
by definition, not smart enough to debug it." -- Brian Kernighan
Craig DeForest
2013-11-21 15:57:58 UTC
Permalink
I don't have a strong reason to push for 5.12 over 5.10. But I'm with you, David, on the need for a policy. "Last five stable versions" seems reasonable.


On Nov 21, 2013, at 7:20 AM, David Mertens <***@gmail.com> wrote:

> That said, I think it might be nice to provide a clear dependency policy. For example, we could promise to support all default Perls on the latest versions of Cygwin, Strawberry, Active Perl, Mac OSX, Debian stable, Centos, and Fedora. Note that some of those---namely Centos and Debian stable---may be a bit conservative and we could instead provide instructions or even install scripts that would install perlbrew and a newer version of Perl.
>
> Or, we could just promise to work on the five latest stable versions of Perl (which, at the moment, would be 5.18, 5.16, 5.14, 5.12, and 5.10), and therefore (try to) nudge Cygwin to move along. If we keep working with the oldest Perl, then the distributors have no need to move forward, right?
>
> David
>
>
> On Thu, Nov 21, 2013 at 9:10 AM, David Mertens <***@gmail.com> wrote:
> I don't see a need to move to 5.12. I think I'll eventually advocate for that because of the lexical keyword API, but I don't have anything for that yet and I think that Zefram typically writes CPAN modules which provide the same C interfaces for older Perls.
>
> The biggest Perl feature I would like to see is user-level pragmatic modules, which come with 5.10. And actually, what I really want is warmings from a module. AFAICT, that's been available since 5.6, so I suppose I/we could have started on that a long time ago. Heh.
>
> David
>
>
> On Wed, Nov 20, 2013 at 7:11 PM, Chris Marshall <***@gmail.com> wrote:
> Without detailed information on who would be affected
> by a change in the required/supported version of perl,
> I would prefer to minimize disruption for PDL users.
> For example, I do know that the older cygwin releases
> used perl 5.10.x and a jump to 5.12 could make them
> not able to use PDL.
>
> That said, if there is a specific need that could be
> addressed by jumping to 5.12.x, that could justify
> the change. Anything in mind---I haven't seen anything
> myself.
>
> All is not bad, if we go to 5.10 support, we can finally
> use 'say' ... :-)
>
> --Chris
>
>
> On Wed, Nov 20, 2013 at 9:19 AM, Craig DeForest
> <***@boulder.swri.edu> wrote:
> > Why not bump to 5.12?
> >
> > (mobile)
> >
> >
> > On Nov 20, 2013, at 4:42 AM, Chris Marshall <***@gmail.com> wrote:
> >
> >> I propose moving to perl 5.10.x as the officially
> >> supported perl version for general PDL development,
> >> effective immediately.
> >>
> >> I don't expect this to affect the legacy PDL users
> >> as they often are using PDL versions back as far
> >> as 2.4.3.
> >>
> >> Comment, discussion, votes?
> >> Chris
> >>
> >> _______________________________________________
> >> PDL-porters mailing list
> >> PDL-***@jach.hawaii.edu
> >> http://mailman.jach.hawaii.edu/mailman/listinfo/pdl-porters
> >>
>
> _______________________________________________
> Perldl mailing list
> ***@jach.hawaii.edu
> http://mailman.jach.hawaii.edu/mailman/listinfo/perldl
>
>
>
> --
> "Debugging is twice as hard as writing the code in the first place.
> Therefore, if you write the code as cleverly as possible, you are,
> by definition, not smart enough to debug it." -- Brian Kernighan
>
>
>
> --
> "Debugging is twice as hard as writing the code in the first place.
> Therefore, if you write the code as cleverly as possible, you are,
> by definition, not smart enough to debug it." -- Brian Kernighan
Judd Taylor
2013-11-21 19:14:44 UTC
Permalink
One of the reasons (well, the only reason, really), that I need 5.8.8 is because that's what RHEL5 is running, and RHEL 5 is supposed to be supported for quite a while. Maybe the policy should be that we should support Perls as old as are in current support for major distributions, especially the ones that are supposed to remain stable for a given timespan, like 5 years in the case of RHEL 5.

What specifically are you planning on introducing into the code that will break older perls than 5.10?

-Judd

____________________________
Judd Taylor
Software Engineer

Orbital Systems, Ltd.
3807 Carbon Rd.
Irving, TX 75038-3415

(972) 915-3669 x127
________________________________
From: Craig DeForest [***@boulder.swri.edu]
Sent: Thursday, November 21, 2013 9:57 AM
To: David Mertens
Cc: Chris Marshall; ***@jach.hawaii.edu; pdl-porters
Subject: Re: [Pdl-porters] [Perldl] update PDL perl version to 5.10.x

I don't have a strong reason to push for 5.12 over 5.10. But I'm with you, David, on the need for a policy. "Last five stable versions" seems reasonable.


On Nov 21, 2013, at 7:20 AM, David Mertens <***@gmail.com<mailto:***@gmail.com>> wrote:

That said, I think it might be nice to provide a clear dependency policy. For example, we could promise to support all default Perls on the latest versions of Cygwin, Strawberry, Active Perl, Mac OSX, Debian stable, Centos, and Fedora. Note that some of those---namely Centos and Debian stable---may be a bit conservative and we could instead provide instructions or even install scripts that would install perlbrew and a newer version of Perl.

Or, we could just promise to work on the five latest stable versions of Perl (which, at the moment, would be 5.18, 5.16, 5.14, 5.12, and 5.10), and therefore (try to) nudge Cygwin to move along. If we keep working with the oldest Perl, then the distributors have no need to move forward, right?

David


On Thu, Nov 21, 2013 at 9:10 AM, David Mertens <***@gmail.com<mailto:***@gmail.com>> wrote:
I don't see a need to move to 5.12. I think I'll eventually advocate for that because of the lexical keyword API, but I don't have anything for that yet and I think that Zefram typically writes CPAN modules which provide the same C interfaces for older Perls.

The biggest Perl feature I would like to see is user-level pragmatic modules<http://perldoc.perl.org/perlpragma.html>, which come with 5.10. And actually, what I really want is warmings from a module<http://perldoc.perl.org/perllexwarn.html#Reporting-Warnings-from-a-Module>. AFAICT, that's been available since 5.6, so I suppose I/we could have started on that a long time ago. Heh.

David


On Wed, Nov 20, 2013 at 7:11 PM, Chris Marshall <***@gmail.com<mailto:***@gmail.com>> wrote:
Without detailed information on who would be affected
by a change in the required/supported version of perl,
I would prefer to minimize disruption for PDL users.
For example, I do know that the older cygwin releases
used perl 5.10.x and a jump to 5.12 could make them
not able to use PDL.

That said, if there is a specific need that could be
addressed by jumping to 5.12.x, that could justify
the change. Anything in mind---I haven't seen anything
myself.

All is not bad, if we go to 5.10 support, we can finally
use 'say' ... :-)

--Chris


On Wed, Nov 20, 2013 at 9:19 AM, Craig DeForest
<***@boulder.swri.edu<mailto:***@boulder.swri.edu>> wrote:
> Why not bump to 5.12?
>
> (mobile)
>
>
> On Nov 20, 2013, at 4:42 AM, Chris Marshall <***@gmail.com<mailto:***@gmail.com>> wrote:
>
>> I propose moving to perl 5.10.x as the officially
>> supported perl version for general PDL development,
>> effective immediately.
>>
>> I don't expect this to affect the legacy PDL users
>> as they often are using PDL versions back as far
>> as 2.4.3.
>>
>> Comment, discussion, votes?
>> Chris
>>
>> _______________________________________________
>> PDL-porters mailing list
>> PDL-***@jach.hawaii.edu<mailto:PDL-***@jach.hawaii.edu>
>> http://mailman.jach.hawaii.edu/mailman/listinfo/pdl-porters
>>

_______________________________________________
Perldl mailing list
***@jach.hawaii.edu<mailto:***@jach.hawaii.edu>
http://mailman.jach.hawaii.edu/mailman/listinfo/perldl



--
"Debugging is twice as hard as writing the code in the first place.
Therefore, if you write the code as cleverly as possible, you are,
by definition, not smart enough to debug it." -- Brian Kernighan



--
"Debugging is twice as hard as writing the code in the first place.
Therefore, if you write the code as cleverly as possible, you are,
by definition, not smart enough to debug it." -- Brian Kernighan
Craig DeForest
2013-11-21 20:00:48 UTC
Permalink
It would be nice if we could use the now-ancient // operator, introduced in 5.10.


On Nov 21, 2013, at 12:14 PM, Judd Taylor <***@orbitalsystems.com> wrote:

> One of the reasons (well, the only reason, really), that I need 5.8.8 is because that's what RHEL5 is running, and RHEL 5 is supposed to be supported for quite a while. Maybe the policy should be that we should support Perls as old as are in current support for major distributions, especially the ones that are supposed to remain stable for a given timespan, like 5 years in the case of RHEL 5.
>
> What specifically are you planning on introducing into the code that will break older perls than 5.10?
>
> -Judd
>
> ____________________________
> Judd Taylor
> Software Engineer
>
> Orbital Systems, Ltd.
> 3807 Carbon Rd.
> Irving, TX 75038-3415
>
> (972) 915-3669 x127
> From: Craig DeForest [***@boulder.swri.edu]
> Sent: Thursday, November 21, 2013 9:57 AM
> To: David Mertens
> Cc: Chris Marshall; ***@jach.hawaii.edu; pdl-porters
> Subject: Re: [Pdl-porters] [Perldl] update PDL perl version to 5.10.x
>
> I don't have a strong reason to push for 5.12 over 5.10. But I'm with you, David, on the need for a policy. "Last five stable versions" seems reasonable.
>
>
> On Nov 21, 2013, at 7:20 AM, David Mertens <***@gmail.com> wrote:
>
>> That said, I think it might be nice to provide a clear dependency policy. For example, we could promise to support all default Perls on the latest versions of Cygwin, Strawberry, Active Perl, Mac OSX, Debian stable, Centos, and Fedora. Note that some of those---namely Centos and Debian stable---may be a bit conservative and we could instead provide instructions or even install scripts that would install perlbrew and a newer version of Perl.
>>
>> Or, we could just promise to work on the five latest stable versions of Perl (which, at the moment, would be 5.18, 5.16, 5.14, 5.12, and 5.10), and therefore (try to) nudge Cygwin to move along. If we keep working with the oldest Perl, then the distributors have no need to move forward, right?
>>
>> David
>>
>>
>> On Thu, Nov 21, 2013 at 9:10 AM, David Mertens <***@gmail.com> wrote:
>> I don't see a need to move to 5.12. I think I'll eventually advocate for that because of the lexical keyword API, but I don't have anything for that yet and I think that Zefram typically writes CPAN modules which provide the same C interfaces for older Perls.
>>
>> The biggest Perl feature I would like to see is user-level pragmatic modules, which come with 5.10. And actually, what I really want is warmings from a module. AFAICT, that's been available since 5.6, so I suppose I/we could have started on that a long time ago. Heh.
>>
>> David
>>
>>
>> On Wed, Nov 20, 2013 at 7:11 PM, Chris Marshall <***@gmail.com> wrote:
>> Without detailed information on who would be affected
>> by a change in the required/supported version of perl,
>> I would prefer to minimize disruption for PDL users.
>> For example, I do know that the older cygwin releases
>> used perl 5.10.x and a jump to 5.12 could make them
>> not able to use PDL.
>>
>> That said, if there is a specific need that could be
>> addressed by jumping to 5.12.x, that could justify
>> the change. Anything in mind---I haven't seen anything
>> myself.
>>
>> All is not bad, if we go to 5.10 support, we can finally
>> use 'say' ... :-)
>>
>> --Chris
>>
>>
>> On Wed, Nov 20, 2013 at 9:19 AM, Craig DeForest
>> <***@boulder.swri.edu> wrote:
>> > Why not bump to 5.12?
>> >
>> > (mobile)
>> >
>> >
>> > On Nov 20, 2013, at 4:42 AM, Chris Marshall <***@gmail.com> wrote:
>> >
>> >> I propose moving to perl 5.10.x as the officially
>> >> supported perl version for general PDL development,
>> >> effective immediately.
>> >>
>> >> I don't expect this to affect the legacy PDL users
>> >> as they often are using PDL versions back as far
>> >> as 2.4.3.
>> >>
>> >> Comment, discussion, votes?
>> >> Chris
>> >>
>> >> _______________________________________________
>> >> PDL-porters mailing list
>> >> PDL-***@jach.hawaii.edu
>> >> http://mailman.jach.hawaii.edu/mailman/listinfo/pdl-porters
>> >>
>>
>> _______________________________________________
>> Perldl mailing list
>> ***@jach.hawaii.edu
>> http://mailman.jach.hawaii.edu/mailman/listinfo/perldl
>>
>>
>>
>> --
>> "Debugging is twice as hard as writing the code in the first place.
>> Therefore, if you write the code as cleverly as possible, you are,
>> by definition, not smart enough to debug it." -- Brian Kernighan
>>
>>
>>
>> --
>> "Debugging is twice as hard as writing the code in the first place.
>> Therefore, if you write the code as cleverly as possible, you are,
>> by definition, not smart enough to debug it." -- Brian Kernighan
Tim Jenness
2013-11-21 21:14:33 UTC
Permalink
One of the problems here is that I'm a bit leery of using the system perl
at all for private installations. I stopped doing that years ago as I was
bitten by problems with some bits of the perl coming from rpms and others
from my installs. I now use perlbrew exclusively and my life has much less
pain in it. The decision on what perl PDL should support shouldn't really
be driven by long term maintenance schedules of Linux distributions --
those guys won't be back porting PDL to their distribution anyhow. It's
many years since 5.8.x came out and there are many features in 5.10 and
5.12 that are really useful. One of perl's problems recently has been that
most people haven't even noticed 5.10 and above and think perl is a dead
language. I was also intrigued by the recent discussion on Mavericks and
PDL. I wouldn't dream of using /usr/bin/perl on my Mac.

Note that I can be ignored on all this. I carry no weight.

--
Tim Jenness



On Thu, Nov 21, 2013 at 1:00 PM, Craig DeForest
<***@boulder.swri.edu>wrote:

> It would be nice if we could use the now-ancient // operator, introduced
> in 5.10.
>
>
> On Nov 21, 2013, at 12:14 PM, Judd Taylor <***@orbitalsystems.com>
> wrote:
>
> One of the reasons (well, the only reason, really), that I need 5.8.8 is
> because that's what RHEL5 is running, and RHEL 5 is supposed to be
> supported for quite a while. Maybe the policy should be that we should
> support Perls as old as are in current support for major distributions,
> especially the ones that are supposed to remain stable for a given
> timespan, like 5 years in the case of RHEL 5.
>
> What specifically are you planning on introducing into the code that will
> break older perls than 5.10?
>
> -Judd
>
> ____________________________
> Judd Taylor
> Software Engineer
>
> Orbital Systems, Ltd.
> 3807 Carbon Rd.
> Irving, TX 75038-3415
>
> (972) 915-3669 x127
> ------------------------------
> *From:* Craig DeForest [***@boulder.swri.edu]
> *Sent:* Thursday, November 21, 2013 9:57 AM
> *To:* David Mertens
> *Cc:* Chris Marshall; ***@jach.hawaii.edu; pdl-porters
> *Subject:* Re: [Pdl-porters] [Perldl] update PDL perl version to 5.10.x
>
> I don't have a strong reason to push for 5.12 over 5.10. But I'm with
> you, David, on the need for a policy. "Last five stable versions" seems
> reasonable.
>
>
> On Nov 21, 2013, at 7:20 AM, David Mertens <***@gmail.com>
> wrote:
>
> That said, I think it might be nice to provide a clear dependency policy.
> For example, we could promise to support all default Perls on the latest
> versions of Cygwin, Strawberry, Active Perl, Mac OSX, Debian stable,
> Centos, and Fedora. Note that some of those---namely Centos and Debian
> stable---may be a bit conservative and we could instead provide
> instructions or even install scripts that would install perlbrew and a
> newer version of Perl.
>
> Or, we could just promise to work on the *five* latest stable versions of
> Perl (which, at the moment, would be 5.18, 5.16, 5.14, 5.12, and 5.10), and
> therefore (try to) nudge Cygwin to move along. If we keep working with the
> oldest Perl, then the distributors have no need to move forward, right?
>
> David
>
>
> On Thu, Nov 21, 2013 at 9:10 AM, David Mertens <***@gmail.com>
> wrote:
>
>> I don't see a need to move to 5.12. I think I'll eventually advocate for
>> that because of the lexical keyword API, but I don't have anything for that
>> yet and I think that Zefram typically writes CPAN modules which provide the
>> same C interfaces for older Perls.
>>
>> The biggest Perl feature I would like to see is user-level pragmatic
>> modules <http://perldoc.perl.org/perlpragma.html>, which come with 5.10.
>> And actually, what I really want is warmings from a module<http://perldoc.perl.org/perllexwarn.html#Reporting-Warnings-from-a-Module>.
>> AFAICT, that's been available since 5.6, so I suppose I/we could have
>> started on that a long time ago. Heh.
>>
>> David
>>
>>
>> On Wed, Nov 20, 2013 at 7:11 PM, Chris Marshall <***@gmail.com>
>> wrote:
>>
>>> Without detailed information on who would be affected
>>> by a change in the required/supported version of perl,
>>> I would prefer to minimize disruption for PDL users.
>>> For example, I do know that the older cygwin releases
>>> used perl 5.10.x and a jump to 5.12 could make them
>>> not able to use PDL.
>>>
>>> That said, if there is a specific need that could be
>>> addressed by jumping to 5.12.x, that could justify
>>> the change. Anything in mind---I haven't seen anything
>>> myself.
>>>
>>> All is not bad, if we go to 5.10 support, we can finally
>>> use 'say' ... :-)
>>>
>>> --Chris
>>>
>>>
>>> On Wed, Nov 20, 2013 at 9:19 AM, Craig DeForest
>>> <***@boulder.swri.edu> wrote:
>>> > Why not bump to 5.12?
>>> >
>>> > (mobile)
>>> >
>>> >
>>> > On Nov 20, 2013, at 4:42 AM, Chris Marshall <***@gmail.com>
>>> wrote:
>>> >
>>> >> I propose moving to perl 5.10.x as the officially
>>> >> supported perl version for general PDL development,
>>> >> effective immediately.
>>> >>
>>> >> I don't expect this to affect the legacy PDL users
>>> >> as they often are using PDL versions back as far
>>> >> as 2.4.3.
>>> >>
>>> >> Comment, discussion, votes?
>>> >> Chris
>>> >>
>>> >> _______________________________________________
>>> >> PDL-porters mailing list
>>> >> PDL-***@jach.hawaii.edu
>>> >> http://mailman.jach.hawaii.edu/mailman/listinfo/pdl-porters
>>> >>
>>>
>>> _______________________________________________
>>> Perldl mailing list
>>> ***@jach.hawaii.edu
>>> http://mailman.jach.hawaii.edu/mailman/listinfo/perldl
>>>
>>
>>
>>
>> --
>> "Debugging is twice as hard as writing the code in the first place.
>> Therefore, if you write the code as cleverly as possible, you are,
>> by definition, not smart enough to debug it." -- Brian Kernighan
>>
>
>
>
> --
> "Debugging is twice as hard as writing the code in the first place.
> Therefore, if you write the code as cleverly as possible, you are,
> by definition, not smart enough to debug it." -- Brian Kernighan
>
>
>
> _______________________________________________
> PDL-porters mailing list
> PDL-***@jach.hawaii.edu
> http://mailman.jach.hawaii.edu/mailman/listinfo/pdl-porters
>
>
Matthew Kenworthy
2013-11-21 21:34:21 UTC
Permalink
Hi Tim,

I'm rebuilding the SciPDL against /usr/bin/perl, as it's the least bad (and
most consistently there) version of Perl on a Mac.

Having the ability to choose a perl to build PDL and its dependencies
against would be wonderful (and it's been suggested at ~ 1 year intervals),
but it's not going to happen with me - I'm going through the struggle of
statically linking PGPLOT into PDL, and it's taken me the best part of a
day to remind myself of all that horror. I know you weren't suggesting
that, but I thought I'd nix that suggestion before it grew too far :)

I remember Tim Pickering talking about writing a homebrew
script/package/doohickey, so that may be the better way to go, if it
exists. And of course there's the very unsatisfying RTFM, which is of no
help getting PDL adopters. SciPDL has its warts (no PLplot for starters)
but at least it allows you to run the Book examples.

Cheers,

Matt


On Thu, Nov 21, 2013 at 10:14 PM, Tim Jenness <***@gmail.com> wrote:

> One of the problems here is that I'm a bit leery of using the system perl
> at all for private installations. I stopped doing that years ago as I was
> bitten by problems with some bits of the perl coming from rpms and others
> from my installs. I now use perlbrew exclusively and my life has much less
> pain in it. The decision on what perl PDL should support shouldn't really
> be driven by long term maintenance schedules of Linux distributions --
> those guys won't be back porting PDL to their distribution anyhow. It's
> many years since 5.8.x came out and there are many features in 5.10 and
> 5.12 that are really useful. One of perl's problems recently has been that
> most people haven't even noticed 5.10 and above and think perl is a dead
> language. I was also intrigued by the recent discussion on Mavericks and
> PDL. I wouldn't dream of using /usr/bin/perl on my Mac.
>
> Note that I can be ignored on all this. I carry no weight.
>
> --
> Tim Jenness
>
s***@optusnet.com.au
2013-11-21 22:40:58 UTC
Permalink
From: Judd Taylor
Sent: Friday, November 22, 2013 6:14 AM

> What specifically are you planning on introducing into the code that will
> break older perls than 5.10?

Hi Judd,

There's currently (as of PDL-2.007_01) stuff like "pack 'l<l<A16L<L', ..."
and "pack 'L<*', ..." in PDL/IO/Storable.pm. If that section of code gets
called on pre-5.10, it's a runtime error because those earlier perls don't
understand the coercion to small-endian order via the '<' template.

Perl 5.8 does have "V" which packs an unsigned long into little-endian
order, but I don't see anything that packs a *signed* long into
little-endian order.
It would seem (to me) a pity to have to work around such deficiencies,
though I'm not passionate about this.

Cheers,
Rob
David Mertens
2013-11-21 23:34:06 UTC
Permalink
I don't much care what version we choose. I use much newer vintages of Perl
in everything I do. That said, here are my thoughts:

I agree with Tim's sentiment: I don't think we should care one way or the
other about Linux because perlbrew is just so darn easy. I just proposed
tying ourselves to OS-supplied minima to get ideas rolling. As far as I'm
concerned, anybody on a Linux that uses an ancient Perl should be
encouraged to use the PDL supplied in their package management system or
switch to perlbrew. If they want stability and don't want perlbrew, they
get stability, bugs and all. The perlbrew argument applies equally to Mac
(although Mac does not provide a package management system with PDL).

That leaves us with concerns for Windows users, chiefly Cygwin users. Given
that Chris, our Pumpking, uses Cygwin and has put enormous effort into
getting Windows and Cygwin compatibility, I would strongly oppose chucking
out Cygwin compatibility. That said, Cygwin reportedly ships
5.14<http://cygwin.com/ml/cygwin-announce/2012-07/msg00011.html>.
Also, it may be possible to run perlbrew on
Cygwin<http://stackoverflow.com/questions/12094783/does-perlbrew-work-with-cygwin>.
So, honestly, I see no reason we don't switch to 5.14.

I can't imagine many people having their own scripts or modules break
because they upgraded to a newer Perl.

Given that perlbrew does give some trouble installing on Cygwin, I think we
should tie ourselves to the minimum supported Perl on Cygwin. Reini has
indicated it will be a little while before he moves to a higher version, so
that'll be stable for a little while.

David


On Thu, Nov 21, 2013 at 5:40 PM, <***@optusnet.com.au> wrote:

>
> From: Judd Taylor
> Sent: Friday, November 22, 2013 6:14 AM
>
>
> What specifically are you planning on introducing into the code that will
>> break older perls than 5.10?
>>
>
> Hi Judd,
>
> There's currently (as of PDL-2.007_01) stuff like "pack 'l<l<A16L<L', ..."
> and "pack 'L<*', ..." in PDL/IO/Storable.pm. If that section of code gets
> called on pre-5.10, it's a runtime error because those earlier perls don't
> understand the coercion to small-endian order via the '<' template.
>
> Perl 5.8 does have "V" which packs an unsigned long into little-endian
> order, but I don't see anything that packs a *signed* long into
> little-endian order.
> It would seem (to me) a pity to have to work around such deficiencies,
> though I'm not passionate about this.
>
> Cheers,
> Rob
>
>
>
> _______________________________________________
> PDL-porters mailing list
> PDL-***@jach.hawaii.edu
> http://mailman.jach.hawaii.edu/mailman/listinfo/pdl-porters
>



--
"Debugging is twice as hard as writing the code in the first place.
Therefore, if you write the code as cleverly as possible, you are,
by definition, not smart enough to debug it." -- Brian Kernighan
Judd Taylor
2013-11-21 23:41:35 UTC
Permalink
This is not good news for me, or my 30+ customer systems in the field running 5.8.8 :(

Granted, at the moment, I'm not running the bleeding edge PDL, but should some new feature become a necessity, I would have to find a new OS, and then upgrade everybody (which usually incurs $$ as shipping hard drives with preinstalled OSes is easier than having customers do it).

What would your suggestion be for someone like me as to a workable upgrade path?

As far as pack goes, I've always just chained pack calls together to get what I want. As I do a lot of satellite comm stuff, I tend to use pack and unpack quite a bit to extract satellite data. The only time I needed to resort to PDL::PP and do it in C, was dealing with a satellite that was sending 10bit words (which was a headache, to say the least).

However, I'm not a user of PDL::IO::Storable, so I guess I'm ok with this being broken, as long as no other part of PDL that I'm using is using PDL::IO::Storable in turn.

-Judd


____________________________
Judd Taylor
Software Engineer

Orbital Systems, Ltd.
3807 Carbon Rd.
Irving, TX 75038-3415

(972) 915-3669 x127

________________________________________
From: ***@optusnet.com.au [***@optusnet.com.au]
Sent: Thursday, November 21, 2013 4:40 PM
To: Judd Taylor
Cc: ***@jach.hawaii.edu; pdl-porters
Subject: Re: [Pdl-porters] [Perldl] update PDL perl version to 5.10.x

From: Judd Taylor
Sent: Friday, November 22, 2013 6:14 AM

> What specifically are you planning on introducing into the code that will
> break older perls than 5.10?

Hi Judd,

There's currently (as of PDL-2.007_01) stuff like "pack 'l<l<A16L<L', ..."
and "pack 'L<*', ..." in PDL/IO/Storable.pm. If that section of code gets
called on pre-5.10, it's a runtime error because those earlier perls don't
understand the coercion to small-endian order via the '<' template.

Perl 5.8 does have "V" which packs an unsigned long into little-endian
order, but I don't see anything that packs a *signed* long into
little-endian order.
It would seem (to me) a pity to have to work around such deficiencies,
though I'm not passionate about this.

Cheers,
Rob
David Mertens
2013-11-21 23:58:50 UTC
Permalink
Judd -

I hear your concern loud and clear but requirements for RHEL5 could live
out to 2017, or even
2020<https://access.redhat.com/site/support/policy/updates/errata/#Life_Cycle_Dates>.
I don't think it's sensible to tie ourselves to that. Obviously changing
the cutting-edge PDL won't change previous versions, and so older PDLs
running on your machines will run without trouble. Furthermore---if you can
make a strong case for needing a newer PDL on your customers' machines---I
would rather volunteer to help you build a custom rpm for PDL with redacted
sections, rather than hold PDL back from climbing to newer Perls. I might
even entertain the notion of helping build an rpm that installs a newer
Perl to /usr/local. And yes, that is a serious offer. I would hate to
unnecessarily drop support for your clients, but I would also hate to hold
PDL back because RHEL5 will hold on to a version of Perl that will be over
a decade out of official support.

David


On Thu, Nov 21, 2013 at 6:41 PM, Judd Taylor <***@orbitalsystems.com>wrote:

> This is not good news for me, or my 30+ customer systems in the field
> running 5.8.8 :(
>
> Granted, at the moment, I'm not running the bleeding edge PDL, but should
> some new feature become a necessity, I would have to find a new OS, and
> then upgrade everybody (which usually incurs $$ as shipping hard drives
> with preinstalled OSes is easier than having customers do it).
>
> What would your suggestion be for someone like me as to a workable upgrade
> path?
>
> As far as pack goes, I've always just chained pack calls together to get
> what I want. As I do a lot of satellite comm stuff, I tend to use pack and
> unpack quite a bit to extract satellite data. The only time I needed to
> resort to PDL::PP and do it in C, was dealing with a satellite that was
> sending 10bit words (which was a headache, to say the least).
>
> However, I'm not a user of PDL::IO::Storable, so I guess I'm ok with this
> being broken, as long as no other part of PDL that I'm using is using
> PDL::IO::Storable in turn.
>
> -Judd
>
>
> ____________________________
> Judd Taylor
> Software Engineer
>
> Orbital Systems, Ltd.
> 3807 Carbon Rd.
> Irving, TX 75038-3415
>
> (972) 915-3669 x127
>
> ________________________________________
> From: ***@optusnet.com.au [***@optusnet.com.au]
> Sent: Thursday, November 21, 2013 4:40 PM
> To: Judd Taylor
> Cc: ***@jach.hawaii.edu; pdl-porters
> Subject: Re: [Pdl-porters] [Perldl] update PDL perl version to 5.10.x
>
> From: Judd Taylor
> Sent: Friday, November 22, 2013 6:14 AM
>
> > What specifically are you planning on introducing into the code that will
> > break older perls than 5.10?
>
> Hi Judd,
>
> There's currently (as of PDL-2.007_01) stuff like "pack 'l<l<A16L<L', ..."
> and "pack 'L<*', ..." in PDL/IO/Storable.pm. If that section of code gets
> called on pre-5.10, it's a runtime error because those earlier perls don't
> understand the coercion to small-endian order via the '<' template.
>
> Perl 5.8 does have "V" which packs an unsigned long into little-endian
> order, but I don't see anything that packs a *signed* long into
> little-endian order.
> It would seem (to me) a pity to have to work around such deficiencies,
> though I'm not passionate about this.
>
> Cheers,
> Rob
>
>
> _______________________________________________
> PDL-porters mailing list
> PDL-***@jach.hawaii.edu
> http://mailman.jach.hawaii.edu/mailman/listinfo/pdl-porters
>



--
"Debugging is twice as hard as writing the code in the first place.
Therefore, if you write the code as cleverly as possible, you are,
by definition, not smart enough to debug it." -- Brian Kernighan
Dima Kogan
2013-11-26 06:43:36 UTC
Permalink
***@optusnet.com.au writes:

> From: Judd Taylor
> Sent: Friday, November 22, 2013 6:14 AM
>
>> What specifically are you planning on introducing into the code that will
>> break older perls than 5.10?
>
> Hi Judd,
>
> There's currently (as of PDL-2.007_01) stuff like "pack 'l<l<A16L<L', ..."
> and "pack 'L<*', ..." in PDL/IO/Storable.pm. If that section of code gets
> called on pre-5.10, it's a runtime error because those earlier perls don't
> understand the coercion to small-endian order via the '<' template.
>
> Perl 5.8 does have "V" which packs an unsigned long into little-endian
> order, but I don't see anything that packs a *signed* long into
> little-endian order.
> It would seem (to me) a pity to have to work around such deficiencies,
> though I'm not passionate about this.

For the record, I don't think supporting perl 5.8 is worth the effort.
This particular incompatibility is trivial to deal with, however, so I
just pushed a fix. I have no perl 5.8 to test with, so this is untested.
The '<' pack symbols are no longer used, and if that's the only issue,
it'll work in 5.8 now.

Note that I also removed the 5.10 check from the test suite.

dima
Loading...