Windows-x86 (32-bit) build is broken. Time to retire it?

[Magnus Ihse Bursie]

On 2023-12-06 14:22, Stewart Addison wrote:

> I can understand that, however we should bear in mind that with 
> JDK<=17 being supported regardless, if it gets removed in a later 
> version there is almost certainly a much higher risk of backports 
> "breaking" the LTS streams (as it did recently) compared to other 
> platforms without it being caught prior to integration, so it might be 
> good to catch it earlier in the backport cycle (or maybe as a middle 
> ground just add such checks to JDK<=17) regardless of whether it gets 
> removed in >=21.

That's just a specific case of the general cost of doing backports: 
stuff change in mainline, so you can never know how much effort it will 
be to make the corresponding change in an older release. That can not 
be, and have never been, a valid argument for blocking changes in mainline.

The current policy of the update project seem to be a very much "open 
arms", accepting a lot of issues for backport, even if they are not 
critical for the stability or the security of the platform. Such a 
policy is of course more likely to run into problems the more the 
mainline diverges from the version the update release is based on. But 
once again, that weight is on the update maintainers. If it becomes too 
much risk of breakage when backporting changes from mainline, then they 
might need to lower their acceptance rate. Once again, it should not 
affect what decisions we make for the mainline going forward.

/Magnus


>
> Regards,
>
> Stewart...
> --
>
>
>
> On Wed, 6 Dec 2023 at 11:54, Alan Bateman <Alan.Bateman at oracle.com> wrote:
>
>     On 06/12/2023 10:22, Stewart Addison wrote:
>     > Hi Magnus,
>     >
>     > I put in a fix for another Win32 build break recently but hadn't
>     > realised it was broken again - ref
>     > https://bugs.openjdk.org/browse/JDK-8319958. Since the Win32
>     port is
>     > currently only deprecated I was already wondering if we should
>     go for
>     > the simple solution of having some github PR checks added for Win32
>     > along with the other platforms (Obviously it would need the
>     parameter
>     > to enable deprecated ports) as I feel this would help prevent
>     things
>     > being merged and subsequently backported that break things.
>
>     I assume this would put a tax on everyone, the testing in GHA is
>     already
>     a bit noisy. Seems like it would be better to remove the port, I
>     think
>     it would be useful to hear from George Adams and Bruno Borges to
>     know if
>     they plan to follow-up JEP 449 and do the next step.
>
>     -Alan
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://mail.openjdk.org/pipermail/jdk-dev/attachments/20231207/cd26f9fb/attachment.htm>