Ok, it seems that I'was not convinced about the "safty" of "&aOut->out"
but it seems
that (at least gcc) add the aOut into the first item of "method_base".
-> This is infact the "frienddef" I required by "default" :-)
so - it seems it was my fould. and the tool https://godbolt.org/z/e_z8_y
include compile test
helped.
On 03.02.20 11:02, J Decker wrote:
On Mon, Feb 3, 2020 at 1:38 AM aotto <aotto1968@xxxxxxxxxxx
<mailto:aotto1968@xxxxxxxxxxx>> wrote:
Hi, as i mentioned earlier… the "not having" create a couple of
problems:
1. the "void-pointer-inflation"
you can rewrite "method_base(struct base* hdl, ...)" win
"method_base(void* hdl, ...)" this seems to solve the problem
BUT you definitely loos the type-check because other pointer as
"A" and "B"
also allowed.
2. the "cast inflation"
you can add the "CPP" wrapper to "method_base" this will act
like "void*"
but you loose the "type-check also"
3. even if you use "&aObj->obj" this create EXTRA CODE which add
only CRASH-scenarios
and add EXTRA overhead…
3. there is only ONE clear solution which NOT loose the
"type-checking" and DON'T add
extra code and this solution is the "frienddef"
None of these are an issue, unless you're actually trying to go the
other direction (getting an A* for B* from a base*), which none of
your examples are exemplifiying....
1) nothing prevents you from passing the address of the base structure
2) I think you mean #define wrapper; which can still have your
in-library type check for 'is a valid X'
3) does not create extra code, since your base pointer is the base of
your structure, it'll still be NULL, unlike other places... you've
already constrained it that it's the first member.
3(2)) since the rest of the points aren't themselves valid it doesn't
appear there really needs to be an additional solution...
mfg AO
On 03.02.20 10:20, J Decker wrote:
On Mon, Feb 3, 2020 at 1:10 AM aotto <aotto1968@xxxxxxxxxxx
<mailto:aotto1968@xxxxxxxxxxx>> wrote:
On 03.02.20 10:07, J Decker wrote:
> On Mon, Feb 3, 2020 at 12:23 AM aotto
<aotto1968@xxxxxxxxxxx <mailto:aotto1968@xxxxxxxxxxx>> wrote:
>
>> Hi,
>>
>> there is an easy test, if Stefan is right and I'm wrong.
>>
>> 1. cast an INVALID Pointer to "aObj"
>>
> how would friend prevent an invalid pointer pass?
→ I have a pointer-type-and-valid test in my software, this mean
I know if a pointer is "valid" or not... but this works
only if
I get the pointer into my "base_method".
C doesn't have any such thing, no do any other developer, so how
would it help them? (you fell off the list)
