RFR: 8324683: Unify AttachListener code for Posix platforms [v3]
Martin Doerr
mdoerr at openjdk.org
Fri Apr 12 10:17:44 UTC 2024
On Fri, 12 Apr 2024 05:49:45 GMT, Thomas Stuefe <stuefe at openjdk.org> wrote:
> > Maybe you could take my stuff and consolidate it with yours so we can get closer to the number of lines deleted that my PR has, or even beat it :-)?
> > All the best, Johan
>
> Hi @jdksjolen,
>
> Duplicate work is always annoying, sorry. To our defense, @SoniaZaldana 's work has been going on for many weeks now out in the open.
>
> I advised her to leave AIX out of the patch for several reasons:
>
> When analyzing the differences, we found a number of sublteties that may break AIX, and we have no easy way to check that platform. From my time as AIX porter I remember that the attach listener framework on AIX was exceptionally brittle and took a lot of grooming to get right.
>
> Since the current AIX porters are really overworked atm and the last thing they need are regressions, I would leave the AIX code base as it is for now. Let the porters deal with this if/when they find time. Importantly, they then also can double-check whether many of the old workarounds we added back in the early 2000's (eg those close/shutdown workarounds) are still needed in modern AIX versions.
>
> Or, what do you think @TheRealMDoerr?
>
> Therefore, I would rather see Sonia's patch in its current form go in.
AIX work in this area is typically done by @JoKern65 or @varada1110. Maybe you would like to take a look.
It's probably ok to do AIX later, but I can't tell if the new code will make that harder or easier.
Is the code in attachListener_posix.cpp all specified by posix or just a collection of things which are used on many posix based operating systems?
-------------
PR Comment: https://git.openjdk.org/jdk/pull/18283#issuecomment-2051473592
More information about the hotspot-runtime-dev
mailing list