[bug#38509] [PATCH] gnu: libuv: Update to 1.34.0

[Andrew Miloradovsky]

Hi Brett,

WRT the differences, it seems to be mostly fixes and refactoring:

- https://github.com/libuv/libuv/tree/v1.34.0

WRT the failed tests, it is due to treating EAGAIN as the failure:

- https://github.com/libuv/libuv/issues/2531

Not sure how to properly fix it, fixing the tests themselves is
non-trivial, and retrying several more times likely won't work either.

On 12/7/19 1:34 AM, Brett Gilio wrote:
> Brett Gilio <address@hidden> writes:
>
>> Andrew Miloradovsky <address@hidden> writes:
>>
>>> -    (arguments
>>> -     '(;; XXX: Some tests want /dev/tty, attempt to make connections, etc.
>>> -       #:tests? #f))
>>> +    (arguments '(#:tests? #f))
>>> +    ;; tests 122-124 (getnameinfo_basic_ip*) fail
>>> +    ;; https://github.com/libuv/libuv/issues/2531
>> Hi Andrew,
>>
>> Thank you for your submission. I have some questions before we go
>> forward with this patch. First, I'd like to note that this change would
>> trigger a huge rebuild of thousands of dependent packages.
>>
>> brettg@oryx ~/Repos/guix$ ./pre-inst-env guix refresh -l libuv
>> Building the following 2067 packages would ensure 5397 dependent packages 
>> are rebuilt:
>>
>> That said, is there some particular functionality or security that is
>> provided with this patch? If so, it would be great if you could
>> elaborate on that in the Git sub-header of the commit message.
>>
>> Since this patch would trigger such a massive rebuild it will need to go
>> to the `core-updates` branch to rest before it sees master. So having
>> that detail in the commit message will make it easier for us to see what
>> we are working with.
>>
>> Lastly, just curious if there is a way to work around the issues with
>> tests 122-124 as shown in your above snippet. I know that the tests were
>> blanket disabled before, so I am just curious if there is a _better_ way
>> to do this, maybe i'm wrong. If the issue is that the tests require
>> network functionality that can usually be spoofed with some
>> effort. Also, stylistically, comments not on the same line as the code
>> they are commenting usually go before the code in question.
>>
>> I hope that makes sense! Let me know if you have any questions, I am
>> happy to help.
> Actually, on further exploration of the `core-updates` branch, libuv is
> already at 1.34 there.
>