As explained elsewhere, rust selftest is going to be disabled to
resolve sporadic autobuilder failures. This also gives everyone a
chance to update rust to latest upstream release without having to
fight with the selftest.

Before selftest is re-enabled, I would ask for the following improvements:

- when it fails, the output is huge, and easily exceeds the terminal
buffer. It's difficult to look through it for possible failures, and
impossible to tell if some failures have scrolled off the buffer. Is
it written to a file somewhere? How do the fails look like? The test
should make both of these points clear. Can we reduce verbosity by
default? No one needs to see a complete list of everything that passed
when just a few items fail.

- the test takes 20 minutes to execute. When debugging it you can't
possibly ask for that much time for developers for every test
execution. There should be a way to re-run just the item that failed
from command line in a preserved build-st directory, and that should
be documented and printed on failure.

Thanks,
Alex

On Tue, 14 May 2024 at 17:52, Randy MacLeod via lists.openembedded.org
<[email protected]> wrote:
>
> On 2024-04-05 7:28 a.m., Alex Kiernan via lists.openembedded.org wrote:
>
> On Thu, Apr 4, 2024 at 10:07 AM Yash Shinde via lists.openembedded.org
> <[email protected]> wrote:
>
> On Wed, Apr 3, 2024 at 05:51 PM, Alex Kiernan wrote:
>
> Hi Sundeep (or anyone else with insight on this!)
>
> How do you go about debugging rust oe-selftest failures? I've bumped
> everything up to 1.77 (https://github.com/akiernan/poky) but
> oe-selftest fails for reasons that aren't immediately obvious to me...
>
> https://gist.github.com/akiernan/6cc6131c1ec3af866098a9318679cf1b
>
> Any clues?
>
> --
> Alex Kiernan
>
>
>
>
> Hi Alex,
>
> I faced the same errors when I was working on to upgrade rust from 1.75 to 
> 1.76. I discussed it with rust upstream here, 
> https://github.com/rust-lang/rust/issues/122075 and the following change 
> solves those failures,
>
> https://github.com/rust-lang/rust/issues/122075#issuecomment-1985153524
>
> PR- 
> https://github.com/rust-lang/rust/pull/122205/commits/5aece7fad06baaa745784d118db862b3e3ccf7f8
>
>
>
>
>
> After this change was backported to poky, the rust build was successful but 
> it was seen that not all the tests are executed.
>
> While discussing upstream, it was known that removing the --doc option from 
> python3 src/bootstrap/bootstrap.py test --exclude<some excluded test cases 
> list> --doc --no-fail-fast --bless --target x86_64-poky-linux-gnu  cmd in 
> oeqa/selftest/cases/rust.py file execute the other tests also.
>
>
>
>
>
> But, while doing so another set of failures are observed as mentioned in 
> https://github.com/rust-lang/rust/issues/122285
>
> Complete log using --doc option-  
> https://gist.github.com/Yashinde145/7db6bd4a064021bf756b0c0dd6c8777c
>
> Complete log without using --doc option-  
> https://gist.github.com/Yashinde145/036a934f0523307859f7c855b83ecfd6 (this 
> file is truncated in gist due to large size, please see Complete raw file)
>
> I am not very familiar to these failures but  trying to understand them.
>
> Feels like you're way ahead of me. I'll have a bit more of a poke at
> it when I've some time, but it feels like we may be mostly waiting on
> upstream (and a big +1 for getting them engaged!)
>
> FYI,
>
> Sundeep and Yashe have Rust 1.77 (8?) compiling reproducibly.
> They found some problems with the rust test suite and
> are working on fixing or avoiding those errors before submitting a patch.
>
>
> ../Randy
>
>
>
>
>
>
>
> --
> # Randy MacLeod
> # Wind River Linux
>
>
> 
>
-=-=-=-=-=-=-=-=-=-=-=-
Links: You receive all messages sent to this group.
View/Reply Online (#199447): 
https://lists.openembedded.org/g/openembedded-core/message/199447
Mute This Topic: https://lists.openembedded.org/mt/105307188/21656
Group Owner: [email protected]
Unsubscribe: https://lists.openembedded.org/g/openembedded-core/unsub 
[[email protected]]
-=-=-=-=-=-=-=-=-=-=-=-

Reply via email to