Re: [PATCH 12/13] dlfcn,elf: implement dlmem() [BZ #11767]

[stsp <stsp2@yandex.ru>]

29.03.2023 19:35, Carlos O'Donell пишет:
> The fact that such binaries can be created is enough for me to raise
> a sustained objection to the inclusion of dlmem().

Then someone needs to tell how exactly
it can be created, and, more importantly,
point in the implementation the place
that will then break.
I still am quite sure we have some horrible
misunderstanding, but that misunderstanding
is very unusual as I can't sense its roots.

All elf segments are arranged to their vaddr's
by the elf loader, not by the user. This is
obvious, undisputable fact. You always tell
me I want to lend that task to the user.
No I don't.
Why do you think I do?

>> If you can't, then you go away.
>> Do you accept that challenge?
> Asking a developer to go away not the way consensus is built.

I meant to say "after giving an ACK". :)
OK, sorry if it happened to be a bad joke.
But somehow I need to motivate Jonathon
or anyone else to make a proof of the statements
we have here. Otherwise we can't proceed.

I still leave the second part of a challenge:
if it is demonstrated that my patch doesn't
arrange the segments per vaddr's, I'll go
away forever. v9 has a few bugs in that area,
but not even nearly as bad as to not relocate
segments by vaddr's.


> If consensus can't be reached then the API will not be included
> in glibc.
>
> The real issue for me is that we are not providing a way for developers
> to manage the complexity of the *setup* that is required before dlmem()
> can operate reliably.

No, its not the case.
Its not the case and its not the case.
And I don't know what to do, if I can't
even ask someone to demonstrate why
its the case. :)