Ken Williams wrote:
On Oct 8, 2006, at 10:50 PM, Adam Kennedy wrote:
o I now think Adam mistook my comments as personal criticism and
over-reacted. My apologies if I caused the problem, but I fear Adam
is hyper-sensitive since Data::package is one of his
I'll admit I take some issue to call
On Oct 8, 2006, at 10:50 PM, Adam Kennedy wrote:
o I now think Adam mistook my comments as personal criticism and
over-reacted. My apologies if I caused the problem, but I fear
Adam is hyper-sensitive since Data::package is one of his
I'll admit I take some issue to calling a perfectly nor
On Oct 8, 2006, at 5:49 PM, Ron Savage wrote:
o Just for the record, at $work, I checked several machines and
they all had 2
versions of Perl installed, but in each case I only checked the
compiler used
for the Perl I use, and in each case it was absent. This is a PITA,
of course,
and since
On Oct 8, 2006, at 10:59 PM, Adam Kennedy wrote:
The only downside of this approach is that it requires a CPAN
module for every external system. This is probably quite reasonable
for the top 20 cases and for complicated cases (such as "do I have
a C/XS/etc compiler") but might not be ideal
On Oct 8, 2006, at 11:32 PM, Adam Kennedy wrote:
It seems to me that if a given distro can be built either with a
compiler or without one (e.g. Params::Validate), then the answer
to "does it require a compiler" is a simple "no".
Only if that "either" is across all platforms.
If something,
On Oct 9, 2006, at 1:39 PM, Andreas J. Koenig wrote:
Recent CPAN shells have the "failed" command that sums up which
problems have been encountered in the current session.
Per default a summary of the problems during every command is printed
before the next prompt.
And by turning on color you
On Tue, 10 Oct 2006 01:13:17 +1000, Adam Kennedy wrote:
Hi Adam
> If this is all going to be held in some external system outside the
> META.yml , which is what read the above as going with now, then by
> all means. This is all perfectly workable and fine, changing it
> down the track is just fin
> On Sun, 08 Oct 2006 20:35:29 -0700, Michael G Schwern <[EMAIL PROTECTED]>
> said:
> That's a problem which can be solved by the CPAN shells through
> better error reporting and its not unique to this issue.
Recent CPAN shells have the "failed" command that sums up which
problems
Ron Savage wrote:
On Sun, 8 Oct 2006 16:58:50 -0700, Eric Wilhelm wrote:
Hi Eric
How about doing this with an external tool (i.e. a utility or
Sure, that's OK.
What info will it put in the meta data?
I can't say exactly, but I'm thinking that some general principles could be
enunciated
Adam Kennedy wrote:
>> The only thing left is to make it clear to authors that they must add
>> CBuilder to build_requires if their distro uses M::B features that
>> need it. Documentation is the weak remedy. An automatic error during
>> "./Build dist" is the strong remedy.
>
> Or take the M:I a
On Sun, 08 Oct 2006 17:54:55 -0700, Michael G Schwern wrote:
Hi Michael
> Here's an alternative approach. Why make compiler dependencies
> special? Treat them like any other alien dependency. How do we do
> that? Wrap a module around it. ExtUtils::CBuilder is one example.
> The various Alie
On Sun, 8 Oct 2006 16:58:50 -0700, Eric Wilhelm wrote:
Hi Eric
> How about doing this with an external tool (i.e. a utility or
Sure, that's OK.
> What info will it put in the meta data?
I can't say exactly, but I'm thinking that some general principles could be
enunciated /and/ followed:
o S
# from Marvin Humphrey
# on Sunday 08 October 2006 11:31 pm:
>> An automatic error during "./Build dist" is the strong remedy.
>To follow the automagic-add approach out ad absurdum, should it ?
>automatically detect that you need other things besides CBuilder?
Maybe. But, we're talking about co
On Oct 8, 2006, at 10:26 PM, Eric Wilhelm wrote:
An automatic error during "./Build dist" is the strong remedy.
Aside from the fact that I am apparently unable to correctly configure
my mail client, what is wrong with this approach?
http://www.nntp.perl.org/group/perl.module.build/380
Fro
# from Marvin Humphrey
# on Sunday 08 October 2006 08:13 pm:
>An automatic error during "./Build dist" is the strong remedy.
Aside from the fact that I am apparently unable to correctly configure
my mail client, what is wrong with this approach?
http://www.nntp.perl.org/group/perl.module.buil
It seems to me that if a given distro can be built either with a
compiler or without one (e.g. Params::Validate), then the answer to
"does it require a compiler" is a simple "no".
Only if that "either" is across all platforms.
If something, for example, was looking for /proc/ files on Linux, b
On Oct 8, 2006, at 5:54 PM, Michael G Schwern wrote:
Here's an alternative approach. Why make compiler dependencies
special? Treat them like any other alien dependency. How do we do
that? Wrap a module around it. ExtUtils::CBuilder is one example.
I like. Adding ExtUtils::CBuilder as
A key of "needs_compiler" raises the question, "for what language?" "Compiler"
does not necessarily mean "C compiler". PDL likes to have a Fortran compiler,
Inline::Java needs a Java compiler, and so on.
So the value would be a list of languages.
# Any ol C compiler will do
compiler => 'C
Here's an alternative approach. Why make compiler dependencies special? Treat
them like any other alien dependency. How do we do that? Wrap a module around
it. ExtUtils::CBuilder is one example. The various Alien modules are another.
There's already existing wrapper modules for ssh, rsync
# from Ron Savage
# on Sunday 08 October 2006 03:49 pm:
>o I am thinking of some procedure the author, perhaps, could use to
> generate the info to add to the meta data, like so:
>(1) Probe for compiler
>(2) If found, remove compiler's directory from the path, and bugger
> the consequences (if onl
On Sun, 8 Oct 2006 14:41:49 -0500, Ken Williams wrote:
Hi Ken, and Adam
Thanx for your comments.
Having slept on it, I'd like now to recast this as a tragedy (joke) in 3.14159
acts...
Part 1: The story so far
I've tried to understand Adam's viewpoint, and here's what I assume happened:
o I ha
On Oct 7, 2006, at 9:09 PM, Ron Savage wrote:
Hi Ken, Adam
I had a slight suspicion when I first posted that this might be a
can of worms,
and now I know :-(.
Nah, it's not too bad. We can at least solve the problem I'm
suggesting we solve, which I think was your original request too.
On Oct 7, 2006, at 9:28 PM, Adam Kennedy wrote:
Which is why I'm suggesting that a basic boolean flag isn't quite
good enough, because it doesn't deal with installers (and there's
quite a number of them) that can optionally make use of a compiler
to influence their dependencies.
I know t
On Sat, 7 Oct 2006 20:32:19 -0500, Ken Williams wrote:
Hi Ken, Adam
I had a slight suspicion when I first posted that this might be a can of worms,
and now I know :-(.
--
Cheers
Ron Savage, [EMAIL PROTECTED] on 8/10/2006
http://savage.net.au/index.html
Let the record show: Microsoft is not an Aus
On Sat, 07 Oct 2006 22:55:12 +1000, Adam Kennedy wrote:
Hi Adam
> Yet, because the author has a C compiler installed, what will go
> into the META.yml file will be a dep on YAML::Syck. If you then
> recurse to YAML::Syck, it will say it needs a C compiler, and your
recurse?
> "do I need a C com
On Oct 7, 2006, at 8:05 PM, Adam Kennedy wrote:
From the META.yml side, assuming it's possible to add something to
that that isn't totally useless (because not only does it suffer
from the advisory-quality-only issue, but it does so recursively as
well) you want a flag that isn't boolean.
Ken Williams wrote:
On Oct 5, 2006, at 5:50 AM, Ron Savage wrote:
Of course I don't expect Data::Package to tell me that a dependency
requires
compilation, although dare I say that should be the goal of this first
step.
Yeah, I agree. Probably this will be best handled by the higher-level
On Oct 5, 2006, at 5:50 AM, Ron Savage wrote:
Of course I don't expect Data::Package to tell me that a dependency
requires
compilation, although dare I say that should be the goal of this
first step.
Yeah, I agree. Probably this will be best handled by the higher-
level tools like CPAN a
On Fri, 06 Oct 2006 09:14:04 +1000, Adam Kennedy wrote:
Hi Adam
> One of the problems with this idea of adding the can_cc() data to
> the META.yml is that most authors will have them, and so even if
> the Makefile|Build.PL files have a fallback option for people
> without compilers, that will not
Hi Folks
I'd like to see the META doc extended to tell me that a module requires a
compiler.
For instance, this week I tried installing Data::Package (purely for data -
yeah, right), and the dependency list - eventually - led to Scalar::List::Utils,
which requires compilation, but doesn't tell yo
30 matches
Mail list logo
