[
https://issues.apache.org/jira/browse/TINKERPOP-3030?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17813739#comment-17813739
]
ASF GitHub Bot commented on TINKERPOP-3030:
-------------------------------------------
FlorianHockmann commented on PR #2468:
URL: https://github.com/apache/tinkerpop/pull/2468#issuecomment-1924148227
Just to clarify: This PR doesn't change which frameworks we support with
Gremlin.Net. It's only about which frameworks we use to build and test it.
Gremlin.Net mainly [targets .NET Standard
2.0](https://github.com/apache/tinkerpop/blob/master/gremlin-dotnet/src/Gremlin.Net/Gremlin.Net.csproj#L21)
which is implemented by various .NET frameworks like .NET 5, .NET 6, .NET 7,
and .NET 8. It has .NET 6 as an additional explicit target because we
optionally use some APIs which were only introduced in .NET 6 to support
Websocket compression. This can therefore only be used by users on .NET 6 and
higher.
Before this PR, we built and tested Gremlin.Net with .NET 6. With this PR,
we're using .NET 8 instead. But users can still stay on .NET 6 (or 7 or
whatever) and use newer versions of Gremlin.Net with it.
Not sure if this was clear to you and you just wanted to propose that we
**test** against .NET 6 **and** .NET 8. In that case, I agree with you in
general that it would be good to test on both frameworks (or better to say
runtimes in this case). The problem with that however is that all contributors
then also need both runtimes installed. If you just have .NET 8 installed, then
the tests will fail as you also need .NET 6 in that case.
That's also why we always use the same version across all release branches.
In the end, the question comes down to how many contributors are actually
affected by this and how often do we expect to find version specific bugs
because of it. I guess many contributors simply don't execute .NET tests on
their machines and leave that up to GH Actions or they use the Docker based
setup to execute the tests. So, this probably doesn't affect many contributors.
On the other hand,
[TINKERPOP-3029](https://issues.apache.org/jira/browse/TINKERPOP-3029) is also
the only bug that comes to my mind which we could have catched by executing
tests on different runtimes. Although I think it wouldn't have helped much as
we still would have to be on .NET 8 in the first place to find it. So, the
better question is how often a bug occurs that is specific to an **older**
framework than the latest LTS version which we want to use by default.
I don't have a strong opinion on this. If we are confident that this doesn't
really affect contributors negatively, then I'm in favor of executing the tests
on multiple runtimes. Otherwise, I'd stay with our current approach.
> Update to .NET 8
> ----------------
>
> Key: TINKERPOP-3030
> URL: https://issues.apache.org/jira/browse/TINKERPOP-3030
> Project: TinkerPop
> Issue Type: Improvement
> Components: dotnet
> Affects Versions: 3.5.8, 3.6.6, 3.7.1
> Reporter: Florian Hockmann
> Assignee: Florian Hockmann
> Priority: Minor
>
> .NET 8 is now the latest LTS release so we should update to it.
> We need to do that on all release branches so contributors only need to have
> one version of .NET installed on their systems.
> This also means that we test by default with .NET 8 which would have shown
> problems like the one described in TINKERPOP-3029.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
