So my opinion is to release Pekko 1.0.1 as soon as we can. Even though this
was picked up in Scala 3.3.2 RC (The Scala team added incubator-pekko to
their community build which basically means incubator-pekko gets compiled
against the latest Scala version on a nightly basis), the undefined
behaviour is for any version of Scala 2/Scala 3, i.e. reading the Scala 2
spec from
https://www.scala-lang.org/files/archive/spec/2.11/01-lexical-syntax.html

> The ‘$’ character is reserved for compiler-synthesized identifiers. User
programs should not define identifiers which contain ‘$’ characters.

This means hypothetically any future change to Scala compiler (inclusive of
Scala 3 which is how the problem got detected in the first place) can
create undefined behaviour so it's in our best interest to at least get
Pekko 1.0.1 with this fix out as soon as we can. That being said, since the
affected code is private for Pekko there could also practically be zero
issues so I don't think it's worth it to delay a pekko-http release over
this (although its somewhat annoying because it does mean pekko-http will
get released with a "broken" version of pekko and we would then ideally
want to make another future pekko-http release quickly using the fixed
pekko 1.0.1). On this point I would wait until tomorrow to see what the
Scala compiler team says and how critical it is, if it's deemed very
critical then delaying pekko-http until pekko 1.0.1 is released to me makes
sense (but irregardless if we get a response or not we should ideally start
releasing the fixed pekko 1.0.1 tomorrow).

Regards

On Mon, Jul 17, 2023 at 4:32 PM PJ Fanning <[email protected]> wrote:

> Hi everyone,
>
> Due to a class name issue [1], we are planning to produce an RC1 for
> Pekko 1.0.1 over the next day or two.
>
> So far the issue appears in compatibility testing for Scala 3.3.2-RC1
> and hasn't been reported elsewhere. The class in question is in
> pekko-remote.
>
> The code is package private so I don't think it will cause major
> issues when we get 1.0.1 released (i.e. users switching from v1.0.0 to
> v1.0.1 are very unlikely to have any issues relating to us changing
> the problematic class name).
>
> There is nothing in the `main` branch yet that we wouldn't want to
> include in the v1.0.1 release so we will make the RC from the `main`
> branch.
>
> I still think we can also press on with a Pekko HTTP 1.0.0-RC1 in the
> coming days. We might delay it by a couple of days but I think we can
> release Pekko HTTP 1.0.0 with a dependency on Pekko (Core) 1.0.0 but
> release note that users should use Pekko (Core) 1.0.1 when it is
> available.
>
> The alternative is to delay Pekko HTTP 1.0.0-RC1 until Pekko (Core)
> 1.0.1 is released. With the current 2 phase voting requirements and
> the build process that we have, that probably means delaying Pekko
> HTTP 1.0.0-RC1 by 2 weeks.
>
> Feel free to share your opinions on this thread.
>
> Regards,
> PJ
>
> [1] https://github.com/apache/incubator-pekko/issues/491
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: [email protected]
> For additional commands, e-mail: [email protected]
>
>

-- 

Matthew de Detrich

*Aiven Deutschland GmbH*

Immanuelkirchstraße 26, 10405 Berlin

Amtsgericht Charlottenburg, HRB 209739 B

Geschäftsführer: Oskari Saarenmaa & Hannu Valtonen

*m:* +491603708037

*w:* aiven.io *e:* [email protected]

Reply via email to