Ruby 3.3 compatible release?

Ruby 3.3 compatible release?

[Samuel Williams]

Hi Eric,

Thanks for your continued effort to maintain and support this gem - despite how old it is - I assume you were not expecting it to still be in use today!

We are trying to clean up the `io.h` interface in Ruby 3.3 and would like to know if there is anything we can do to help with a release of kgio which supports the updated interface and removes usage of deprecated symbols. That is because it is still a dependency of Unicorn, which, as of today has a market share of around 15% - something I hope you are really proud of!

I believe Jean Boussier has already generously supplied a patch:

https://github.com/casperisfine/kgio/commit/a04f6057a94ae8413fc6d1e7b74bfa6b9d802285.patch

However, it does not appear to be merged or released yet. Anything we can do to help move things forward? Would it help if I resubmitted it directly using git format-patch?

Warm regards,
Samuel

Re: Ruby 3.3 compatible release?

[Eric Wong]

Samuel Williams <samuel@oriontransfer.net> wrote:
> Hi Eric,
 
> Thanks for your continued effort to maintain and support this
> gem - despite how old it is - I assume you were not expecting
> it to still be in use today!
> 
> We are trying to clean up the `io.h` interface in Ruby 3.3 and
> would like to know if there is anything we can do to help with
> a release of kgio which supports the updated interface and
> removes usage of deprecated symbols. That is because it is
> still a dependency of Unicorn, which, as of today has a market
> share of around 15% - something I hope you are really proud
> of!

Hello Samuel, please refrain from positive comments about unicorn+kgio
as they make me extremely uncomfortable.

Not only am I a shy introvert, I'm nearly certain the unintentional
popularity of these projects damaged the entire Ruby+Rack ecosystem and
set it back ~15 years in parallelism, concurrency and robustness
compared other languages+runtimes by allowing users to ignore many
types of bugs.  The negatives outweight any positives, here.


Back to the release:

Are there any other incompatibilities that may happen for Ruby 3.3
that we might have to deal with before December?  I don't want to
have to scramble another release together if we make one too soon.


Again, I'm a shy introvert and have been struggling to even make
releases+announcements for more important projects that keep me fed.


New releases also increases download counts on RubyGems.org which puts
them in danger of MFA requirements, so I try to limit them nowadays.
MFA is more work for me to setup since I refuse to deal with corporate
Terms of Service.  I never wanted people to trust me, not just because
I'm shy but also due to the irreparable damage I've done to Ruby,
copyleft, and centralization[2].

> I believe Jean Boussier has already generously supplied a patch:
> 
> casperisfine/kgio/commit/a04f6057a94ae8413fc6d1e7b74bfa6b9d802285.patch

> However, it does not appear to be merged or released yet.
> Anything we can do to help move things forward? Would it help
> if I resubmitted it directly using git format-patch?

I forgot about it originally because I was struggling with SSD errors
and eventual failure.  Given Jean's email was neither created with
"git format-patch" nor "git request-pull" it didn't get picked up in
later search queries, either.

Even seeing the name of that hosting platform is continual reminder
of my failures[1][2]

So please resend with "git format-patch".  Thanks.



[1] I used to tell people that "all git hosts are created equal"
    when their UI was merely Embrace+Extend.  They're now actively
    Extinguishing copyleft licenses with AI and I can't think of a way
    the Free Software community could fight it given the monetary and
    legal resources of the other side.


[2] Fwiw, I hate centralization so I created git-svn in 2006 to kill
    SVN, only to have a larger, more powerful centralized monster
    replace it.  So not only am I to blame for damaging Ruby+Rack
    with unicorn, I'm also largely to blame for the early popularity
    of said centralized platform because SVN migrants weren't ready
    for centralization-resistant workflows.  FML :<

Re: Ruby 3.3 compatible release?

[Samuel Williams]

> On 28/08/2023, at 12:58 PM, Eric Wong <e@80x24.org> wrote:
> 
> Are there any other incompatibilities that may happen for Ruby 3.3
> that we might have to deal with before December?  I don't want to
> have to scramble another release together if we make one too soon.

As far as I know, and have tested, there are no other incompatibilities, and I think the Ruby team has a high standard for avoiding such situations. It’s partly my fault for pushing forward with changes, but we didn’t know what incompatibilities, if any, would occur due to the changes. Ruby 3.3, w.r.t `io.h`, should be fully backwards compatible, with deprecations.

We just merged a PR in Ruby head which amounts to:

#define HAVE_RB_IO_T

This restores compatibility with kgio and unicorn.

That being said, I think it would be great to merge the appropriate fixes which remove usage of deprecated fields in Ruby 3.3.

I’ll follow up with `git format-patch` soon.

Warm regards,
Samuel