[csw-pkgsubmissions] newpkgs libruby18_1, ruby, ruby18, ruby18_dev(...)

[Philip Brown]

On Fri, Mar 18, 2011 at 5:37 PM, Ben Walton <bwalton at opencsw.org> wrote:
>
>> And you have told me that you are going to ignore that fact, simply
>> because it is *my* site?
>
> No, it has nothing to do with what site you're at, Phil.  It's the
> fact that you've taken an issue that wasn't even on the radar until
> three days ago when I raised a question on the list to now block a
> package because it's not done in the way you prefer it to be.

Yes, its true that if you didnt raise the question on the list, I
probably would not have noticed this issue.
I dont notice EVERYTHING about ALL packages.. I'm not omniscient. But
when my attention *is* drawn to an issue, at least I care enough to
look at it. That should be a good thing?

and I will repeat what I have said previously: My concern is not that
it is done *my* way; my concern is that it is done in the way that is
*best for our users*.
If it is shown that your way is best for our users, I will heartily
add my endorsement to it and send the package on its way, even though
"my way" lost.

So, I'm still waiting on some feedback from someone outside the two of us.
Hopefully we'll see some after the weekend.


> [1] If you look for ENV['CC'] and expect an unqualified binary to make
>    the choice, you're opening the issue to wider problems.  You're
>    then going to assume that cc is sun pro when it need not be.

I think assuming "cc = sun cc" is a reasonable assumption on a Solaris machine.
it is, after all, the same assumption that pretty much every single
autoconf makes on a solaris machine these days.

> Same
>    for gcc.  Do you assume it's version 4?  The options in the other
>    variables currently assume that's the case, but what if this
>    changes.  The current mechanisms are using known working
>    combinations of settings.

sure, there is room for a bit of vagueness.. but it's durn tootin
sure, that if CC=gcc, then passing in gcc flags, is going to be a
whole lot better than passing in Sun cc ones.

Appending a reply to your follow-on email:

>The rbconfig.rb file contains _nothing_ but a toggle based on
>ENV['CC'].
>
>If rbconfig.ENV['CC'].rb exists, require it.  If ENV['CC'] is nil or
>rbconfig.ENV['CC'].rb doesn't exist, require rbconfig.default.rb.
>Alternatives should then be used to toggle rbconfig.default.rb between
>the available options.

Thats a very forward thinking idea, that has some merit to it. If you
want to take what I have suggested as a patch, "own" it, and add in
your own suggestion above, please do so.
It is, unfortunately, beyond my own level of ruby skills to maintain
such a thing.



PS: I think it is reasonable to put in a limit for your package, that
says somewhere, "If you plan to use gcc with our ruby packages, we
only currently support gcc v4"