Bug#982175: #982175 task-japanese-desktop: should explicitly prefer mozc over anthy
Adding debian-japanese@l.d.o to the loop
> Package: task-japanese-desktop
> Version: 3.63
> Tags: bullseye patch
> Priority: important
>
> Dear Maintainer,
>
> Most Japanese users prefer "mozc" over "anthy" among several Japanese
> input method engines.
> Because mozc is supported upstream in limited architectures (x86 and
> arm related only)
> anthy has been used as a fall-back.
> However, the recent bullseye debian installer does not configure the
> system as expected.
>
> When the buster (or earlier) debian installer installs japanese-desktop task
> it configures (runs postinst maintainer scripts of) the uim-* input
> method packages in the following order:
> uim-anthy
> uim-mozc
>
> buster dpkg.log:
> 2021-01-31 12:47:57 install uim-anthy:all <なし> 1:1.8.8-4+deb10u3
> 2021-01-31 12:47:57 status half-installed uim-anthy:all 1:1.8.8-4+deb10u3
> 2021-01-31 12:47:57 status unpacked uim-anthy:all 1:1.8.8-4+deb10u3
> (snip)
> 2021-01-31 12:51:44 install uim-mozc:amd64 <なし> 2.23.2815.102+dfsg-4
> 2021-01-31 12:51:44 status half-installed uim-mozc:amd64 2.23.2815.102+dfsg-4
> 2021-01-31 12:51:44 status unpacked uim-mozc:amd64 2.23.2815.102+dfsg-4
> (snip)
> 2021-01-31 12:53:58 configure uim-anthy:all 1:1.8.8-4+deb10u3 <なし>
> 2021-01-31 12:53:58 status unpacked uim-anthy:all 1:1.8.8-4+deb10u3
> 2021-01-31 12:53:58 status half-configured uim-anthy:all 1:1.8.8-4+deb10u3
> 2021-01-31 12:53:58 status installed uim-anthy:all 1:1.8.8-4+deb10u3
> (snip)
> 2021-01-31 12:54:04 configure uim-mozc:amd64 2.23.2815.102+dfsg-4 <なし>
> 2021-01-31 12:54:04 status unpacked uim-mozc:amd64 2.23.2815.102+dfsg-4
> 2021-01-31 12:54:04 status half-configured uim-mozc:amd64 2.23.2815.102+dfsg-4
> 2021-01-31 12:54:04 status installed uim-mozc:amd64 2.23.2815.102+dfsg-4
>
> which makes uim input method framework configured in the desirable
> preference ordering,
> because the uim-* input method packages are structured as LIFO: last
> configured package is preferred first.
>
> However, the recent bullseye debian installer (perhaps recent apt or
> dpkg algorithm) has been changed for some reason. It configures the
> uim-* input method packages in the following order:
> uim-mozc
> uim-anthy
>
> bullseye dpkg.log:
> 2021-01-31 10:55:44 install uim-anthy:all <none> 1:1.8.8-7
> 2021-01-31 10:55:44 status half-installed uim-anthy:all 1:1.8.8-7
> 2021-01-31 10:55:44 status unpacked uim-anthy:all 1:1.8.8-7
> (snip)
> 2021-01-31 10:58:47 install uim-mozc:amd64 <none> 2.26.4220.100+dfsg-2
> 2021-01-31 10:58:47 status half-installed uim-mozc:amd64 2.26.4220.100+dfsg-2
> 2021-01-31 10:58:47 status unpacked uim-mozc:amd64 2.26.4220.100+dfsg-2
> (snip)
> 2021-01-31 11:00:36 configure uim-mozc:amd64 2.26.4220.100+dfsg-2 <none>
> 2021-01-31 11:00:36 status unpacked uim-mozc:amd64 2.26.4220.100+dfsg-2
> 2021-01-31 11:00:36 status half-configured uim-mozc:amd64 2.26.4220.100+dfsg-2
> 2021-01-31 11:00:36 status installed uim-mozc:amd64 2.26.4220.100+dfsg-2
> (snip)
> 2021-01-31 11:00:38 configure uim-anthy:all 1:1.8.8-7 <none>
> 2021-01-31 11:00:38 status unpacked uim-anthy:all 1:1.8.8-7
> 2021-01-31 11:00:38 status half-configured uim-anthy:all 1:1.8.8-7
> 2021-01-31 11:00:38 status installed uim-anthy:all 1:1.8.8-7
>
> Since depending on the dpkg configuration order is inherently fragile,
> it should be avoided.
> It should be better to adjust the dependency field in the task package.
> The attached patch should fix this problem.
What do Japanese people think about this?
Would you support prefering mozc over anthy?
Holger
--
Holger Wansing <hwansing@mailbox.org>
PGP-Fingerprint: 496A C6E8 1442 4B34 8508 3529 59F1 87CA 156E B076
Reply to: