GitLab

Like generated po/Makefile.in.in file does, to decide where to
instal the translation files.
Instead of guessing it based on the prefix, causing us not to show
the translations when staring in appropriate locale.
Bug #761372 (m.rick.mac)
And this fixes the 3 test cases that have been failing for me for
a while.

Bug #762728 (Dominique Leuenberger)

This must have been wrong for quite a long time, since GTK+ changed
its expand system.

Bug #762108 (m.rick.mac)

This seems to fix the crash when choosing an image repeatedly:
https://bugzilla.gnome.org/show_bug.cgi?id=761396#c13

Because recent versions of libevince-view have an assert that crashes
the application:
https://bugzilla.gnome.org/show_bug.cgi?id=761396#c12

It seems useful.

I guess this is why I haven't been seeing the busy cursor recently.
Found by clang-tidy.

Doing that just to unconst the members is still awful, but at least
this is explicit about what we are doing.

Why didn't the compiler catch this?

See https://github.com/isocpp/CppCoreGuidelines/blob/master/CppCoreGuidelines.md#sl50-avoid-endl

So we can use std::unique_ptr<> instead of std::shared_ptr<> for the
parser member variable, and generally make the APIs slightly less
tightly bound.

Because nothing shares it.

Found by clang-tidy's misc-use-override check.

With help from clang-modernize-3.7.

With help from clang-modernize-3.7.

Otherwise, we still get the dropped table name when asking libgda
for a list of tables. This caused Glom to complain that a
previously-deleted table name already existed when the user tried to
recreate it.
Bug #754645 (m.rick.mac)