Re: atk,pango,gtk+/cygwin

[Tor Lillqvist <tml iki fi>]

Stefan Ondrejicka writes:
 > * pango/pango-utils.c: use DllMain also on cygwin
 >   pango_get_sysconf_subdirectory(): 
 >   pango_get_lib_subdirectory(): on cygwin optionaly return standard "unix"
 >   path or "win32" path

I am not sure this is a good idea. On Unix, it is normal that binary
distributions of software like GTK+ or Pange contain hardcoded paths,
and that they must be installed in a location chosen by the person who
configured and compiled the package. As Cygwin is supposed to look and
work like Unix, it should be OK to use the Unix hardcoded path
principle in GTK+, Pango etc for Cygwin.

Software distribuuted with Cygwin requires that you have stuff where
the Cygwin packagers want them to be. For instance, groff requires
that the fonts are in /usr/share/groff/1.17.2/font.

I.e., I think it's OK to use the Unix-style "hardcoded paths only"
approach on Cygwin, and not use the Win32 "deduce installation
directory from DLL location" as a fallback. If some end-user *has* to
install something in a non-standard place, for instance because of
disk space issues, he can always set up a symbolic link so that the
stuff still is available from the hardcoded path.

But that's only my feeling, and I am not really a heavy-duty Cygwin
person. It would probably be best to ask on the Cygwin list what
people think. I won't apply that part of your patch until I have had
some more opinions... (Please forward whatever consesus might be
reached on the Cygwin list to me or to this list, I am not on the
Cygwin list.)

--tml