On Thu, 2006-03-09 at 09:58 +0100, Werner Punz wrote:
> Mario Ivankovits schrieb:
> > Hi!
> > 
> > Its again me ... somewhat angry, so sorry *grrrr* I CANT SET A
> > BREAKPOINT IN SHARED ANY MORE - or I have to do it twice in shared_impl
> > and shared_tomahawk.
> > Please lets discuss again why this refactoring is needed - PLEASE !!!
> > 
> >> Now I propose to get rid of all those class.getName() (or
> >> this.getClass().getName where required) and replace them by a string
> >> literal (representing the FQN of the class without any refactoring)
> >>   
> > We cant use the FQN as then the refactoring happen on this string
> > literal anyway.
> > So I change my proposal to use org.apache.myfaces.XXXXXX
> > 
> > WDYT?
> > Mario
> > 
> > 
> +1 for a new discussion, this refactoring is outright problematic,
> what we have here is a precompiler situation which basically kills
> several important low level tools right off.

I'm still very much in favour of *not* shipping a common jar used by
both core and tomahawk (and maybe tobago, ADF, etc); the versioning
issues related to that are nasty.

What I had in mind when this was being discussed was not a compilation
pre-processor but instead a tool based on ASM that would post-process a
jarfile to rename all classes from package X to package Y, and change
all references to stuff in package X to the new package Y. Whether this
approach would actually be better or worse with respect to breakpoints
and similar issues I'm not sure. 

One difference of this approach is that the source code would refer to
org.apache.myfaces.shared.X, not to org.apache.myfaces.shared_zzz.X
which might help. The post-processed binary *would* refer to
shared_zzz.X though which might confuse interactive debuggers worse than
source-level preprocessing.

Personally I don't use breakpoints much; logging works better for me for
debugging.

Cheers,

Simon

Reply via email to