On Mon, 24 Mar 2025 13:41:44 GMT, Zdenek Zambersky <zzamb...@openjdk.org> wrote:

> This fixes stack smashing issue in awt library on linux s390x. (jdk compiled 
> with `-fstack-protector-strong`)
> 
> Fix is based on patch 
> [submitted](https://mail.openjdk.org/pipermail/awt-dev/2019-July/015337.html) 
> in JDK-8227919 review thread, rebased to master. They decided to go for 
> Solaris only fix there, with [expected 
> follow-up](https://mail.openjdk.org/pipermail/awt-dev/2019-July/015347.html) 
> for linux/s390x. But that never happened.
> 
> I was not able to get response from original author for this issue, so I am 
> creating PR myself.
> 
> **Testing:**
> I tested jdk with and without this fix on linux/s390x (using Xvfb). It fixes 
> the issue (reproducer for this bug no longer crashes the JVM). I have also 
> tried to run `jdk_awt` tests, where lot of tests was also affected by this. 
> With this patch amount of failures dropped by ~100. 
> 
> Without fix:
> 
>    TEST                                              TOTAL  PASS  FAIL ERROR  
> SKIP   
>>> jtreg:test/jdk:jdk_awt                             2171  1063   415     6   
>>> 687 <<
> 
> 
> With fix
> 
>    TEST                                              TOTAL  PASS  FAIL ERROR  
> SKIP   
>>> jtreg:test/jdk:jdk_awt                             2171  1162   316     6   
>>> 687 <<
> 
> 
> Many tests are no longer crashing the VM:
> 
> -java/awt/Choice/ChoiceKeyEventReaction/ChoiceKeyEventReaction.java           
>                                            Failed. Unexpected exit from test 
> [exit code: 134]
> +java/awt/Choice/ChoiceKeyEventReaction/ChoiceKeyEventReaction.java           
>                                            Passed. Execution successful
> 
> There seem to be few unstable tests in jdk_awt. (Ones that fail in one or 
> another run, but seem unrelated to this fix)
> 
> I have not done any additional manual testing of XIM as did original author 
> (in review thread mentioned higher), as I am not at all familiar with input 
> method for Japanese (or other Asian characters), or how that supposed to 
> work. (So I cannot verify that input of Asian characters works as expected on 
> linux/s390x, but at least JVM no longer crashes.)

Looks reasonable although it seems an odd behaviour  (bug?) of  XGetICValues

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

Marked as reviewed by prr (Reviewer).

PR Review: https://git.openjdk.org/jdk/pull/24197#pullrequestreview-2823434995

Reply via email to