RFR: 8343618: Stack smashing in awt_InputMethod.c on Linux s390x

[Zdenek Zambersky]

On Tue, 20 May 2025 14:18:06 GMT, Zdenek Zambersky <zzambers at openjdk.org> wrote:

>>> That looks right, and it should also be good for backports. The root cause must be a bug in Xlib, though. Has anyone found it, or reported it?
>> 
>> That was my thought too  : "Looks reasonable although it seems an odd behaviour (bug?) of XGetICValues"
>
>> That looks right, and it should also be good for backports. The root cause must be a bug in Xlib, though. Has anyone found it, or reported it?
> 
> I have not identified the root cause. I think bug could also be in kinput2 or it's interaction with X11/Xlib (as scenario with kinput2 is problematic case as far as I understand).
> 
> Unfortunately this would be very difficult to debug for me, as I have no experience with input of Asian characters (kinput2) and RHEL (8) does not offer kinput2 package. In addition I can only do testing remotely on temporarily loaned s390x headless machine. So I do not plan to dig deeper into this myself and I rely on observations by @takiguc here.

> @zzambers Sorry for my late reaction. Is it possible to add test result by using `XMODIFIERS=@im=local` ?

I have now also tested this scenario (fixed jdk + `XMODIFIERS=@im=local` + Xvfb):
- my reproducer -> ok (no crash)
- jdk_awt tests:

   TEST                                              TOTAL  PASS  FAIL ERROR  SKIP   
>> jtreg:test/jdk:jdk_awt                             2171  1218   260     6   687 <<

No crashes (stack smashing), even less failures (even though lot of changes in failures could probably be attributed to unstable tests... :/ )

-------------

PR Comment: https://git.openjdk.org/jdk/pull/24197#issuecomment-2894683771