Support for older distros

[Geert Janssens]

(Resending because my mail client messed up the text only part of the original 
message and the list server discards the html part...)

On donderdag 23 maart 2017 17:36:18 CET John Ralls wrote:
> > 
> > Is there anything else we changed requirements for ?
> 
> Thanks for working through all of that.
> 
> Yes,  ICU, though we don't require a recent version and its API is very
> stable. There's also Googletest but that's so easy to install from Github
> that I think we shouldn't worry about it.

Ok. Both come prepackaged for Mingw64 by the way.
And FYI, I'm getting this info from here:
https://github.com/Alexpux/MINGW-packages
This is the official repository for packages available via msys2 (which is the 
primary platform to use with mingw64 IIUC)

> 
> There's also gettext (see bug 759844
> <https://bugzilla.gnome.org/show_bug.cgi?id=759844>), but that's more an
> issue for what we use to build release tarballs than anything that users or
> distro packagers need to be concerned about.
> 
Ah yes, the gettext bug. There are two actually the one you mention here and 
Bug 745941 - Review of po Headers / make pot[1] 
The former is about translatable strings being omitted from glade files if 
they have a context attribute *and* we're building in tree. This bug can 
simply be avoided by building out of tree. That however triggers a bug in 
intltool versions older than 0.51 (which incidentally is the last released 
version and rather old by now).

The latter bug started out as a bug reporting two translation issues. One has 
been fixed, the other is actually the very same intltool issue of causing 
absolute paths in gnucash.pot and derived po files. Again the immediate 
solution is to switch to intltool 0.51 and keep building out of tree.

The experiments I reported on that bug all date from before intltool 0.51 got 
released.

Ideally we should dump intltool completely as it's no longer maintained. I did 
a few experiments today to generate our gnucash.pot file using gettext only 
but didn't get it right first time. So this needs some more experimentation. 
Frank's wiki page on this[2] has a very good summary and pointers to start 
from. As for the gettext version to set as baseline, based on Frank's summary 
ideally it should be 0.19.3. However if we stick with intltool, this can be 
relaxed quite a bit as in that case intltool will take care of all that has 
been added in 0.19. In that case RHEL/Centos 7's 0.18.2 is probably sufficient 
already.

> Since MinGW64 supports directly a lot of dependencies that we've been
> building from source--with significant difficulty in some cases--maybe we
> should put some effort into redoing the Windows build system to use that
> instead of vanilla MinGW sooner rather than later.
> 
Agreed. It's been on my todo list for some time now and I hoped to get to it 
before the 2.8 release.

However it looks like I'm stacking projects to work on here and may not have 
enough time to tackle them all. So if others are interested in helping out, 
please chime in :)

Geert

--------
[1] https://bugzilla.gnome.org/show_bug.cgi?id=745941
[2] http://wiki.gnucash.org/wiki/Translation#Background