Fwd: Bug#1091871: libunistring5 1.3-1 causes msgmerge to rewrap a lot of lines

[Santiago Vila]

Hello.

I have received (via Cc) the following report from the Debian bug system.
Since it involves gettext, I'm forwarding it here just in case there
is something that could be advised from the gettext side.

I vaguely remember that in the old times there was a thing
called "canonical po format" or alike. There was really such a thing,
and if so, can we ask msgmerge to output in such format?
(until the standard changes again of course...)

Thanks.

-------- Mensaje reenviado --------
Asunto: Bug#1091871: libunistring5 1.3-1 causes msgmerge to rewrap a lot of 
lines
Resent-Date: Wed, 01 Jan 2025 20:45:01 +0000
Resent-From: Julian Andres Klode <jak@debian.org>
Resent-To: debian-bugs-dist@lists.debian.org
Resent-CC: gettext@packages.debian.org, deity@lists.debian.org, Jörg Frings-Fürst 
<debian@jff.email>
Fecha: Wed, 1 Jan 2025 21:42:16 +0100
De: Julian Andres Klode <jak@debian.org>
Responder a: Julian Andres Klode <jak@debian.org>, 1091871@bugs.debian.org
Para: Debian Bug Tracking System <submit@bugs.debian.org>

Package: libunistring5
Version: 1.3-1
Severity: minor
Control: affects -1 gettext src:apt
X-Debbugs-Cc: gettext@packages.debian.org, deity@lists.debian.org

libunistring5 1.3-1 causes msgmerge to significantly rewrap msgid
and msgstr values, causing a lot of churn in po templates and po
files, for example:

 msgid ""
 -"<!-- TRANSLATOR: used as in -o=config_string e.g. -o=Debug::"
 -"pkgProblemResolver=1 --> <!ENTITY synopsis-config-string
 \"config_string\">"
 +"<!-- TRANSLATOR: used as in -o=config_string e.g. "
 +"-o=Debug::pkgProblemResolver=1 --> <!ENTITY synopsis-config-string "
 +"\"config_string\">"

Downgrading to 1.2-1 fixes it.

To reproduce, run

$ msgmerge --previous po/de.po po/apt-doc.pot  | sponge po/de.po

inside apt's doc directory for a checkout of 2.9.20 and thei

    git diff po/de.po

I feel like this must be unintentional given that the string
being rewrapped in the example above does not contain any
unicode in the first place.
--
debian developer - deb.li/jak | jak-linux.org - free software dev
ubuntu core developer                              i speak de, en