On Tue May 19, 2026 at 7:40 AM CEST, Aditya GS via lists.openembedded.org wrote:
> Hi, Sorry, I forgot to mention the bug reference in my earlier mail.
> This patch is for Bug 16254. This is v2 of the patch that disables the
> randtest in the gawk ptest suite. This version includes the detailed
> debugging analysis showing that the intermittent randtest failures are
> caused by sensitivity in the statistical threshold used in the
> autocorrelation check. Small variations in sigma, floating-point
> behavior, entropy availability, or timing conditions (particularly on
> QEMU / autobuilder) can cause abs(corr[tau]/sigma) to exceed the
> threshold even when the randomness is valid. This aligns with the
> AB-INT failures reported in the autobuilder.
>
> Thanks,
> Aditya GS

Hello,

Thanks for the patch, it looks interesting but:
* Can you follow instruction at [0] to setup git-send-email to send
  properly formatted patches (yours is an attachment whereas the
  supported format is "inline") ?
* Use "[YOCTO #xxx]" to reference the bug you are trying to fix
* Fix you upstream discussion reference (either add the link or remove
  the reference)

[0]: 
https://docs.yoctoproject.org/dev/contributor-guide/submit-changes.html#sending-the-patches-via-email

Thanks!

>
> ________________________________
> From: Aditya GS
> Sent: Tuesday, May 19, 2026 11:06 AM
> To: [email protected] 
> <[email protected]>
> Subject: PATCH gawk: ptest: disable randtest due to nondeterministic behavior
>
> Hi, This is a patch for disabling randtest in the gawk ptest suite. This 
> version includes the debugging analysis.
>
> Thanks,
> Aditya GS


-- 
Yoann Congal
Smile ECS

-=-=-=-=-=-=-=-=-=-=-=-
Links: You receive all messages sent to this group.
View/Reply Online (#237274): 
https://lists.openembedded.org/g/openembedded-core/message/237274
Mute This Topic: https://lists.openembedded.org/mt/119386726/21656
Group Owner: [email protected]
Unsubscribe: https://lists.openembedded.org/g/openembedded-core/unsub 
[[email protected]]
-=-=-=-=-=-=-=-=-=-=-=-

Reply via email to