xgettext incorrectly infers 'c-format' in cpp files

Question

I have a bunch of .cpp files which contain text including a literal percent sign. Those files don't use printf at all, they use iostream stuff:

std::cout <<  _( "50 % of 10 is 5" ) << std::endl;

However, strings containing % are still marked as c-format by xgettext:

#: test.cpp:5
#, c-format
msgid "50 % of 10 is 5"
msgstr ""

This template then gets processed by msgmerge and translated, and translation tools (e.g. Poedit) complains about such entries:

#: test.cpp:5
#, c-format
msgid "50 % of 10 is 5"
msgstr "50 % de 10 est 5"

Format specification in 'msgid' and 'msgstr' does not match

Manually removing the #, c-format comments helps, but they come back after every merge. How can I force xgettext or msgmerge to drop those comments? I know I can just filter them out with sed, but is there a smarter way, which would keep such comments for strings coming from C files?

Answer 1

The gettext documentation talks about this exact same issue here: https://www.gnu.org/software/gettext/manual/html_node/c_002dformat-Flag.html

According to the docs, you should use this special comment before the lines:

xgettext:no-c-format