1 files changed, 39 insertions, 0 deletions
diff --git a/t/fmt-docbook.t b/t/fmt-docbook.t
new file mode 100644
index 0000000..20f0962
--- /dev/null
+++ b/t/fmt-docbook.t
@@ -0,0 +1,39 @@
+# XML and XML-based modules tester.
+
+#########################
+
+use strict;
+use warnings;
+
+use lib q(t);
+use Testhelper;
+
+my @tests;
+
+push @tests,
+  {
+    'format'  => 'docbook',
+    'input'   => 'fmt/docbook/debian-history.dbk',
+    'doc'     => 'Reduced structure of the debian-history document, with the options used in the real package',
+    'options' => "-M UTF-8 -o nodefault='<bookinfo>' -o break='<bookinfo>' -o untranslated='<bookinfo>'",
+  },
+  {    # This is the buggy pre-0.58 behavior where the flow is broken by processing instructions
+     # This triggers an unfortunate behavior as the break occurs on the path <screen><userinput> where <screen> is nowrap and <userinput> is inline|wrap.
+     # As a result, when the PI breaks within <userinput>, the first part is translated with wrap.
+     # When the second part (after the PI) is pushed, we are back in <screen> path, so this part is translated with no-wrap (as it should)
+     # The default behavior was changed to ensure that PI are not breaking, but this option is still provided just in case somebody needs PI to be breaking
+    'doc'     => 'GH#170: processing instructions should not be breaking',
+    'format'  => 'docbook',
+    'options' => '-o break-pi',
+    'input'   => "fmt/docbook/PI-break.dbk",
+  },
+  {    # This is ensuring that GH#170 is gone
+     # Since the PI is handled as inline, the input (that is the same as previously) now builds only one msgid, produced in <screen> path.
+     # The <userinput> is inline as expected, and so is the PI.
+    'doc'    => 'GH#170: processing instructions should be inline tags',
+    'format' => 'docbook',
+    'input'  => "fmt/docbook/PI-inline.dbk",
+  };
+
+run_all_tests(@tests);
+0;