Dne 19. 01. 22 v 9:08 Pavel Valena napsal(a):
On Tue, Jan 18, 2022 at 6:39 PM Vít Ondruch
<vondruch(a)redhat.com> wrote:
>
> Dne 18. 01. 22 v 12:00 Pavel Valena napsal(a):
>> On Mon, Jan 17, 2022 at 4:27 PM Vít Ondruch <vondruch(a)redhat.com> wrote:
>>> Dne 17. 01. 22 v 16:15 Vít Ondruch napsal(a):
>>>> Hi,
>>>>
>>>> It is time of the year when new version of Ruby was released upstream
>>>> and we should land it in Fedora. Unfortunately, the change proposal
>>>> was approved just last Thursday and on top of that, rebase of libffi
>>>> broke Ruby (I am going to disable the failing test cases for the
>>>> moment and hope for the best). So this brings us into situation, where
>>>> won't have enough time prior Fedora Mass rebuild. I have discussed
>>>> this a bit with relengs and one of the options would be to build Ruby
>>>> early during the mass rebuild and fix the outfall later. I shared the
>>>> proposal in the Fedora Mass rebuild ticket [2]. One downside would be
>>>> that in case of problems, we could not trigger our contingency plan,
>>>> which is "drop our side tag". But I hope we won't need
that.
>>>>
>>>> Any thoughts?
>>>>
>>>> My fist concern is that maybe we should build more then just Ruby.
>>>> rubygem-json
>> Hello,
>>
>> I've run the rebuilds of previously failing packages in my COPR
>> (probably with a little older Ruby 3.1 build), and here are the
>> results:
>>
>>
https://copr.fedorainfracloud.org/coprs/pvalena/rubygems-testing/builds/
>> removed other buildroots, so failure marks rawhide buildroot.
>>
>> From that what stands out:
>> 1)
>> ```
>> cannot load such file -- net/smtp
>> ```
>> several packages -
>> e.g.:
https://copr.fedorainfracloud.org/coprs/build/3190889
The other two are rails -related:
rubygem-apipie-rails
rubygem-actionmailbox
rubygem-mail
Thx.
>> 4)
>> ```
>> + ruby -rrubygems -Ilib:test:ext/gio2 test/run-test.rb
>> glib-compile-resources ruby-gio2.gresource.xml
>> cd resource
>> glib-compile-schemas .
>> cd -
>> cd schema/default
>> glib-compile-schemas .
>> cd -
>> cd schema/source
>> cd -
>> Loaded suite test
>> Started
>>
............../usr/share/gems/gems/gobject-introspection-3.4.9/lib/gobject-introspection/loader.rb:616:
>> [BUG] Segmentation fault at 0x00000>
>> ruby 3.1.0dev (2021-12-07 master ec878dac90) [x86_64-linux]
>>
>>
>> ```
>>
https://copr.fedorainfracloud.org/coprs/build/3190970
>
> I hope this is not related to FFI, because gobject-introspection was
> mentioned in that context.
I was not able to rebuilt dependent packages due to this, so the
Could you please finish your thoughts? :)
@Mamoru PTAL, I think that Pavel wanted to say that this is precondition
to build other GTK stuff.
> So probably no blocker, just some backports ....
Agreed. I was checking the build errors randomly from time to time, so
we could hopefully catch those right away :). I won't have time to fix
those failures in the near future though.
What worries me is that I can't rebuild your Ruby PR, still. Do you
have any newer scratch build?
Yes, I sent a link in other subthread:
https://koji.fedoraproject.org/koji/taskinfo?taskID=81413346
In my repo, there is missing this change:
~~~
$ git show 4f6a504c2e9980688763e2164ad0a12c8fdcb9b7
commit 4f6a504c2e9980688763e2164ad0a12c8fdcb9b7
Author: Vít Ondruch <vondruch(a)redhat.com>
Date: Mon Jan 17 17:34:57 2022 +0100
Disable some Fiddle test by libffi-3.4.2.
diff --git a/ruby.spec b/ruby.spec
index bed0ad0..145791c 100644
--- a/ruby.spec
+++ b/ruby.spec
@@ -921,6 +921,11 @@ DISABLE_TESTS="$DISABLE_TESTS -n
!/TestAddressResolve#test_socket_getnameinfo_do
DISABLE_TESTS="$DISABLE_TESTS -n
!/TestReadline#test_interrupt_in_other_thread/"
%endif
+# Several test broken by libffi-3.4.2. There should be fix in libffi, once
+# other components are fixed.
+#
https://bugzilla.redhat.com/show_bug.cgi?id=2040380
+mv test/fiddle/test_import.rb{,.disable}
+
# Give an option to increase the timeout in tests.
#
https://bugs.ruby-lang.org/issues/16921
%{?test_timeout_scale:RUBY_TEST_TIMEOUT_SCALE="%{test_timeout_scale}"} \
~~~
I still have not decided where to push this. Maybe I should push this
even to Rawhide, because otherwise the mass rebuild is going to fail. I
have hoped that this could be already resolved, but it is not :/
Vít