Hi Adhemerval,

13.04.2023 23:09, Adhemerval Zanella Netto пишет:
> It is being tiring to work with your proposal because Rich already
> brought the inherent issue of exposing the loader internals for ELF
> objects [1] about 10 years ago,

My proposal did not exist 10 years ago.
Maybe we all do not properly document
our proposals or an objections. So let me
ask you to please refer to a particular
comment rather than to some 10 years
ago thread, and then carefully document
how can that be the problem in my patches.

I think its a fair ask, e.g. Szabolcs asked
me for a more precise spec - I write. We
progress. Why can't you write the objections
in a detalization level that is enough to
make a progress?

I admit you probably couldn't do that initially,
because I poorly documented my API. But
when Carlos asked for a more detailed spec,
I did. Now you can express your objections
in a detailed manner, let me even attach
the prev API draft to help you doing that.
Of course this draft will be simplified a lot,
but for such a "generic" objections please
use its current form.

>   Szabolcs and myself are constantly
> trying to say that is not good libc interface (regardless it solved
> your specific problem),

And regardless that it was never even discussed.
Part of the failure is on me: I wrote an API draft
only when Carlos asked, I should probably do that
before anything else. Sorry, I am a novice. But its
written now for quite long, so why nothing changes?


>   and even Jonathan tried to explore different
> alternatives.

I am thankful to Jonathon for doing that.
Unfortunately I am unsatisfied with the proposal
of intercepting glibc's internal mmap()s on a
syscall level, and binding a non-trivial (I mean,
very non-trivial!) logic on them. But I am thankful
he proved this is even possible, as I was confident
in an otherwise.

> The main problem is no one from glibc is comfortable, including myself,
> to maintain and work with this interface.

Which no one, besides Szabolcs, have actually
even looked into? It is attached to this e-mail.
Why not to make the complains direct and precise?


>   I don't want to maintain
> another interface with a lot of corner cases.

I have expressed the plans at removing all the
corner cases that Szabolcs pointed to. If you
point more, I'll definitely take an actions.
Off-loading the biggest part to RTLD_NORELOC
will reduce the proposal considerably, avoid
the callback and most of other complexity, so
why such a prejudice? Why can't we just discuss,
amend, see what happens?

> Also the way approaching the glibc community, by not listening to
> criticism and ignoring multiple remarks;

I wonder if Szabolcs also thinks I ignore his
remarks. If you think I ignore yours, then either
that was a misunderstanding on my part, or
they were not referring to my API proposal
(which is attached), or something else from that
line, but definitely not my reluctance to change
the proposal. I want to address every possible
problem raised, but unfortunately so far I only
see the problems raised by Szabolcs.
He details them well and targets over my API
proposals. Its immediately obvious what does
he mean, and after his last review I spent a
full month (!) just thinking how to get rid of
a call-back, and finally got to this RTLD_NORELOC
proposal as a result.


>   saying that you have addresses
> the comments without any ack;

That's a bit strange. Yes, I do some actions
based on a comments I get, then post the
new RFC and write in a log that I addressed
this and that. If there is a further complain,
i can re-do what I thought to be addressed,
and replace it with something I believe now
addresses the comment. Its an iterative
process, misunderstanding happens, etc.
Of course if the policy is to not write "addressed
this and that comment" before getting an
ACK, then I will no longer do that, and write
"tried to address this and that, hope its now better".


>   or flooding the maillist with multiple
> version without addressing previous feedback is really, and I want to
> stress the really here, tiring.

Ok, should I post the next drafts to the bugzilla
then, if flooding ML is bad?
Its not a problem, really!
You tell me what to do, I do.
Why such a policy things are even becoming
a problems? I didn't mean to offend anyone
by flooding the ML, I'll post to bugzilla, and I
don't want you to be offended on me just because
of that. :)