On Sep 19, 2010, at 9:50 AM, Eric MSP Veith wrote:
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
>
> Jeff--
>
> some questions here:
>
> On Sunday 19 September 2010, "Jeff Johnson" <j...@rpm5.org> wrote:
>> + rpmruby ruby = (flags & 0x80000000)
>> + ? rpmrubyI() : rpmrubyGetPool(_rpmrubyPool);
>
> So you found it unncessary to protect the Ruby interpreter from being
> initialized twice (thrice, ...)? That was why I was ripping out this
> portion of the code. (Just asking for some design rationale so that I can
> get it right on the first try next time. ;-))
>
No. The macro syntax always uses the 0x8000000, and I pass
0x8000000 with the scriptlet syntax easily. The global
interpreter (and a single interpreter instance) is all that is being used.
>> #if defined(WITH_RUBYEMBED)
>> - ruby->state = rb_eval_string(str);
>
> You might want to use rb_eval_string_protect to be sure that an exception in
> the Ruby code doesn't kill the whole interpreter and take RPM with it.
>
Noted and tried. But this rule applies to exceptins too:
Never check for an error code you don't know how to handle.
> It seems you have thrown away all of my code. May I ask why? If there was
> some serious design issue or major flaws behind it, please show me. I'm
> always willing to learn (and hesitant to touch code here, anyways). :-)
>
No. I checked into rpm-5_3 to preserve what you checked in on head.
And I was trying to fix a bug, not anything else.
But I don't think ruby 1.9.2 is embeddable atm (I see lots of failures and
no successes after googling).
So perhaps rethinking 1.8.x vs 1.9.x should be done instead?
73 de Jeff
______________________________________________________________________
RPM Package Manager http://rpm5.org
Developer Communication List rpm-devel@rpm5.org
