Re: [OE-core] [PATCH v2 1/1] recipes-support/gnutls: disable slow tests for RISCV64

["Trevor Gamblin" <tgamblin@baylibre.com>]

On Mon Sep 28, 2026 at 3:09 PM EDT, Joao Marcos Costa wrote:
> Hello,
>
> On 9/28/26 7:04 PM, Trevor Gamblin wrote:
>> On Mon Sep 28, 2026 at 12:26 PM EDT, João Marcos Costa wrote:
>>> Timeouts have been observed (very frequently in the past weeks)
>>> in the autobuilder, and a bug [1] was filed at Bugzilla.
>>>
>>> After some investigation [2] was performed by Trevor Gamblin, we now have a
>>> list of the most time consuming test cases:
>>>
>>> - mini-record-2 (187880ms)
>>> - record-retvals (145390ms)
>>> - rng-op-key (63820ms)
>>> - rng-op-random (63330ms)
>>> - rng-op-nonce (63260ms)
>>> - dtls-rehandshake-cert (61360ms)
>>> - dtls-rehandshake-cert-2 (61230ms)
>>> - dtls-rehandshake-anon (61030ms)
>>> - mini-loss-time (60130ms)
>>> - mini-dtls-hello-verify-48 (30050ms)
>>> - tls12-ffdhe (27790ms)
>>> - x509sign-verify-rsa (18460ms)
>>> - x509sign-verify-ecdsa (16050ms)
>>>
>>> Disable them to avoid so many timeouts in qemuriscv64-ptest runs.
>> 
>> Thanks for submitting this. I'm sorry if I wasn't clear before, but I think we
>> should only disable mini-record-2 and record-retvals to start, since they are
>> significantly longer than all of the others. If gnutls remains a problem
>> afterward then the others should be considered, but for now we want to try
>> retaining the others to maximize test coverage.
>> 
>> Trevor
>
> In that case, I'd rather remove at least the cases that take > 60000ms 
> to run, and avoid getting back to this again when the runners are more 
> loaded than usual. What do you think?

Sure, let's do that for now. I'll keep a list of tests disabled for
load/performance reasons and come back to them at some point when we have faster
riscv64 hardware in the AB cluster :)