[Rt-devel] [patch] RT::Interface::Web _Overlay/_Vendor/_Local not being included

Ivan Kohler ivan-rt-devel at 420.am
Tue Apr 26 21:48:23 EDT 2011


On Tue, Apr 26, 2011 at 11:04:52PM +0800, Jesse Vincent wrote:
> 
> On Mon 25.Apr'11 at 17:05:22 -0700, Ivan Kohler wrote:
> > On Thu, Apr 21, 2011 at 06:18:40AM +0400, Ruslan Zakirov wrote:
> > > Hello,
> > > 
> > > This is still doesn't return old behaviour. Which was with side effects.
> > > 
> > > Old code was loading .../Web_Local.pm, but code was compiled in
> > > HTML::Mason::Commands space unless explicit package is defined in
> > > _Local.pm file.
> > 
> > 
> > Okay, how about this instead of the previous patch:
> 
> I believe killing string eval was an explicit goal of the rewrite.


So far, I haven't thought of a way to change the current namespace 
without string eval.  "package" doesn't take a scalar.  Any suggestions?  
Perhaps it can't be done?

Assuming that's the case, how would you like to reconcile the goal of 
killing string eval with Ruslan's request to evaluate the files in the 
HTML::Mason::Commands namespace?


1. Give up on eliminating string eval? (the current patch, probably 
isolated to a Web.pm-only version of _ImportOverlays)

2. Give up on the idea of changing into the caller's current package?
(HTML::Mason::Commands).

3. Some sort of namespace chicanery to evaluate everything in a 
temporary package and then alias it into the desired package?  (without 
string eval)


#2 is my preference - I don't think preserving the "Web_*.pm overlays 
are in the HTML::Mason::Commands namespace" behavior is particularly 
important.

I'd be happy to contribute a patch implementing #1 or #2.

#3 may be more than I can bite off Real Soon, but I guess I could give
it a try if that's the only way you'll even consider fixing the original 
bug that the files aren't pulled in at all...  :)


-- 
Ivan Kohler, President and Head Geek, Freeside Internet Services, Inc.          
Open-source billing, ticketing and provisioning - http://www.freeside.biz/



More information about the rt-devel mailing list