[Pnet-developers] Re: Possible misinterpretation of 'protected internal'

[Carl-Adam Brengesjo]

Dominique Canazzi wrote:
> That must be sorcery!
>
> I posted a fuzzy and quite unreadable report, and it is fixed one day later!
>
> Anyway (and almost for those who found my first sumbission buggy), I 
> attach a better description of the bug, that I have prepared this 
> morning, before it was fixed.
>
> Very, very good job. Many thanks.

check diff to the previous version and you'll notice power of the pnet 
design ;)

> In this case, cscc accepts the code when compiled into a unique assembly (cscc 
> protect*.cs -o protect).
> When I try separate assemblies (cscc -shared -o protect1.dll protect1.cs; cscc 
> protect2.cs -o protect -lprotect1), the compiler outputs:
> protect2.cs:21: called object is not a method or delegate
> protect2.cs:23: called object is not a method or delegate
> In my undersanding, foo() should still be seen as protected, from outside the 
> assembly protect1.
>
> Note that both csc (Microsoft) and mcs (Mono) agree with my interpretation. OK, 
> you may reply that majority does not make truth :-)

Well, it turned out that both you (like Microsoft and Mono) and I was 
correct :-)
I was correct about the ECMA specs, and both csc and mcs does not 
correspond to the ecma spec on this.
However, the ECMA specs, apparently, does not correspond to the 
`conventional oop terms'. So if we're gonna be 'nitty-pitty' we can 
blame the MS guys anyways as they wrote the spec ;)

Glad to learn new stuff anyway :D
(thanks to Gopal that cleared this out for me)

See excerpt from IRC log below 
(http://ajmitch.linuxworks.co.nz/dotgnu.log.08Sep2004).


[23:21] <t3rmin4t0r> FAM OR ASSEM
[23:21] <t3rmin4t0r> FAMILY != conventional meaning of protected
[23:22] <t3rmin4t0r> what ECMA spec says is in fact wrong with the 
accepted meaning of "FAMILY"
[23:23] <t3rmin4t0r> ptha: all protected methods are private to all 
except children *and* nested classes
[23:23] <ptha> so, you're saying that I was right in the interpretation 
of the ecma specs, but not in the sense in conventional oop terms?
[23:23] <t3rmin4t0r> yup

/ptha