Updating json 2.7.5 on F42

[Mamoru TASAKA]

Hello, all:

rubygem-json 2.7.5 is released and I am going to update
rawhide (F42) rubygem-json to this version.
(because the upcoming ruby 3.4.0 will anyway have this or
  higher version of rubygem-json.)

Mass rebuild for this version of rubygem-json shows that
the following pkg(s) will newly FTBFS:

rubygem-activesupport
https://copr.fedorainfracloud.org/coprs/mtasaka/rubygem-newruby-test-6/build...

1) Failure:
TestJSONEncoding#test_hash_keys_encoding [test/json/encoding_test.rb:59]:
--- expected
+++ actual
@@ -1 +1 @@
-"{"\u003c\u003e":"\u003c\u003e"}"
+"{"<>":"\u003c\u003e"}"

git bisect shows that the above failure begins with:
https://github.com/ruby/json/commit/e125072130229e54a651f7b11d7d5a782ae7fb65

So git commit message shows:

'''
Directly use `generate_json_string` for object keys
instead of `generate_json`.

Since the object key is already confirmed to be a string, using a
generic dispatch function brings an unnecessary overhead.
'''

But as the side effect, now object key value is not going to be escaped.
I am not sure if this is desired or not.

Regards,
Mamoru

[Mamoru TASAKA]

Mamoru TASAKA via ruby-sig wrote on 2024/10/31 0:10:
> Hello, all:
>
> rubygem-json 2.7.5 is released and I am going to update
> rawhide (F42) rubygem-json to this version.
> (because the upcoming ruby 3.4.0 will anyway have this or
>   higher version of rubygem-json.)
>
> Mass rebuild for this version of rubygem-json shows that
> the following pkg(s) will newly FTBFS:
>
> rubygem-activesupport
> https://copr.fedorainfracloud.org/coprs/mtasaka/rubygem-newruby-test-6/build...
>
> 1) Failure:
> TestJSONEncoding#test_hash_keys_encoding [test/json/encoding_test.rb:59]:
> --- expected
> +++ actual
> @@ -1 +1 @@
> -"{"\u003c\u003e":"\u003c\u003e"}"
> +"{"<>":"\u003c\u003e"}"
>
> git bisect shows that the above failure begins with:
> https://github.com/ruby/json/commit/e125072130229e54a651f7b11d7d5a782ae7fb65
>
> So git commit message shows:
>
> '''
> Directly use `generate_json_string` for object keys
> instead of `generate_json`.
>
> Since the object key is already confirmed to be a string, using a
> generic dispatch function brings an unnecessary overhead.
> '''
>
> But as the side effect, now object key value is not going to be escaped.
> I am not sure if this is desired or not.

For now reported to rubygem-json:
https://github.com/ruby/json/issues/667

Regards,
Mamoru

[Vít Ondruch]

Hi Mamoru,

On related note, not sure if you have noticed, but there were some code 
updates related to licenses:

https://github.com/ruby/json/issues/277

https://github.com/ruby/json/pull/567

https://github.com/ruby/json/commit/c57d33ec39344f7a6ae2786b8ac36892a51b03fe

So you'd likely want to check the license more carefully this time.

Vít

Dne 30. 10. 24 v 16:10 Mamoru TASAKA via ruby-sig napsal(a):
> Hello, all:
>
> rubygem-json 2.7.5 is released and I am going to update
> rawhide (F42) rubygem-json to this version.
> (because the upcoming ruby 3.4.0 will anyway have this or
>  higher version of rubygem-json.)
>
> Mass rebuild for this version of rubygem-json shows that
> the following pkg(s) will newly FTBFS:
>
> rubygem-activesupport
> https://copr.fedorainfracloud.org/coprs/mtasaka/rubygem-newruby-test-6/build...
>
> 1) Failure:
> TestJSONEncoding#test_hash_keys_encoding [test/json/encoding_test.rb:59]:
> --- expected
> +++ actual
> @@ -1 +1 @@
> -"{"\u003c\u003e":"\u003c\u003e"}"
> +"{"<>":"\u003c\u003e"}"
>
> git bisect shows that the above failure begins with:
> https://github.com/ruby/json/commit/e125072130229e54a651f7b11d7d5a782ae7fb65
>
> So git commit message shows:
>
> '''
> Directly use `generate_json_string` for object keys
> instead of `generate_json`.
>
> Since the object key is already confirmed to be a string, using a
> generic dispatch function brings an unnecessary overhead.
> '''
>
> But as the side effect, now object key value is not going to be escaped.
> I am not sure if this is desired or not.
>
> Regards,
> Mamoru

[Vít Ondruch]

Hi Mamoru,

There have been json 2.7.6 released today. Can I use this as an 
opportunity to ask you to temporary stop updating the json and other 
dual life packages in Rawhide for two reasons:

1) Having those with lower NVR then packages which are coming from 
`ruby` SRPM will be beneficial for Ruby 3.4 rebuild

2) The lower NVR could allow us to reset the release in Ruby (have not 
checked yet if that is feasible).

Nevertheless, neither of those above is must, just nice to have IMHO.

Thx for considering.

Vít

Dne 30. 10. 24 v 16:10 Mamoru TASAKA via ruby-sig napsal(a):
> Hello, all:
>
> rubygem-json 2.7.5 is released and I am going to update
> rawhide (F42) rubygem-json to this version.
> (because the upcoming ruby 3.4.0 will anyway have this or
>  higher version of rubygem-json.)
>
> Mass rebuild for this version of rubygem-json shows that
> the following pkg(s) will newly FTBFS:
>
> rubygem-activesupport
> https://copr.fedorainfracloud.org/coprs/mtasaka/rubygem-newruby-test-6/build...
>
> 1) Failure:
> TestJSONEncoding#test_hash_keys_encoding [test/json/encoding_test.rb:59]:
> --- expected
> +++ actual
> @@ -1 +1 @@
> -"{"\u003c\u003e":"\u003c\u003e"}"
> +"{"<>":"\u003c\u003e"}"
>
> git bisect shows that the above failure begins with:
> https://github.com/ruby/json/commit/e125072130229e54a651f7b11d7d5a782ae7fb65
>
> So git commit message shows:
>
> '''
> Directly use `generate_json_string` for object keys
> instead of `generate_json`.
>
> Since the object key is already confirmed to be a string, using a
> generic dispatch function brings an unnecessary overhead.
> '''
>
> But as the side effect, now object key value is not going to be escaped.
> I am not sure if this is desired or not.
>
> Regards,
> Mamoru