Cross-compilation tools provide the right .pc files and a pkg-config to
match, but often don't provide the scripts to match. Even if they did,
you'd have to change PATH to find them. So give pkg-config precedence.
Signed-off-by: Thiago Macieira <thiago@macieira.org>
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
When that happens, the include/version.h file is in the shadow build
dir, not in the sources.
Signed-off-by: Thiago Macieira <thiago@macieira.org>
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
On Ubuntu and in the source package, libxml2 defaults to naming its
pkg-config file to libxml-2.0.pc so query after the package with that
name to.
I don't know what platform ships it as libxml2, so that might be a typo.
Signed-off-by: Anton Lundin <glance@acc.umu.se>
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
libxslt depends on libxml, so we need xslt before xml on the linker-line.
Signed-off-by: Anton Lundin <glance@acc.umu.se>
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
This requires some serious hacking around with brew to get the libraries and
frameworks that we use to also be able to run on 10.5, but this is the part
that's needed in the Subsurface configuration.
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
On one of my machines codesign doesn't find my signing key by hash, but
does find it by name. Go figure.
Also on that same system (32bit Mac Mini with running Snow Leopard / 10.6)
gcc 4.2 doesn't support the -unused-result warning. So let's only turn
that on for more modern versions of gcc
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
subsurface.pro already useses that.
For consistency with the app, we use swap 4 spaces with
a tab character for indentation.
Signed-off-by: Lubomir I. Ivanov <neolit123@gmail.com>
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
On native Windows the suggested compiler for Qt4.8 is mingw-gcc-4.4.0,
which does not has the -Wno-maybe-uninitialized flag (yet).
Signed-off-by: Lubomir I. Ivanov <neolit123@gmail.com>
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
The *-clang* selector doesn't appear to work correctly in my build environment
(or I just don't understand how it is supposed to work). Either way, making
this conditional on !mac works.
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
Looks like having $$ before QMAKE_HOST.os does not work.
Signed-off-by: Lubomir I. Ivanov <neolit123@gmail.com>
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
Certain versions of mingw32-make will try to "CreateProcess"
passing the contents of CC and CXX. On win32, with CONFIG += silent
CC and CXX become composite containing multiple commands
such as "@echo... && gcc", which produces undesired results.
For the time being, do not automatically enable "silent" mode
if the current OS is win32. Possibly a better solution can be
found later on.
Signed-off-by: Lubomir I. Ivanov <neolit123@gmail.com>
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
Getting dependency information for static libraries requires extra
files. Libtool creates .la files, which are "industry standard". qmake
has its own .prl files, but it can create .la in a pinch if necessary.
This change allows us to get the actual dependency information from
libdivecomputer, without having to guess whether it linked to something
else or not. If it changes in the future, we shouldn't have to do
anything.
Signed-off-by: Thiago Macieira <thiago@macieira.org>
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
Versioning is wrong. It's not a library shared with other projects. And at
least on Mac and Windows we bundle it with the app, anyway.
Requested-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
This is a bit crude of a solution, but the Win32 version
of libxml (or at least some versions) seem to require libiconv,
thus we add the library with pkg-config if it is available.
Signed-off-by: Lubomir I. Ivanov <neolit123@gmail.com>
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
libxml and libxslt use custom -config scripts which brake
qmake and make on Win32. The output of executing a .sh or .cmd
script with the same name produces garbled output. The output
of executables with the same name is untested, but a bit excessive
as a solution.
To work around the issue add another detection stage, which checks
if the previous output from (e.g.) xml2-config is empty and try
to use pkg-config instead.
Signed-off-by: Lubomir I. Ivanov <neolit123@gmail.com>
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
The commented out step should work as expected (in theory).
Signed-off-by: Lubomir I. Ivanov <neolit123@gmail.com>
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
This is a small change, but it seems to work as expected on both
Linux and Win32. What happens if 2> NUL > NUL is used on Win32
for cmd.exe is that the output of both stdout and stderr seems
to pipe into NUL and pkg-config is always reported as missing.
Signed-off-by: Lubomir I. Ivanov <neolit123@gmail.com>
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
In previous build system we linked static to libdivecomputer, so start
doing that again.
Signed-off-by: Anton Lundin <glance@acc.umu.se>
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
We don't use them in our code and Qt doesn't throw either, so save a
few bytes and maybe a few setjump() calls on Windows.
Signed-off-by: Thiago Macieira <thiago@macieira.org>
We're getting a ton of them and they're mostly harmless. I've already
turned on the ones that are problematic (with -Werror even).
Signed-off-by: Thiago Macieira <thiago@macieira.org>
This is working for me, but requires a bit more testing. To build,
run:
qmake [options]
Where options might be:
V=1 disable "silent" build
LIBDCDEVEL=1 use side-by-side libdivecomputer
INCLUDEPATH+=xxx add -Ixxx (e.g., INCLUDEPATH+=/usr/local/marble/include)
LIBS+=xxx add xxx to the linker flags (e.g. LIBS+=-L/usr/local/marble/lib)
or any other qmake option, including debug and release options
If your distribution is already using qtchooser in place of qmake, you
may need to pass an extra option to qmake to select the a
cross-build. For example:
qmake -qt=i686-w64-mingw32-qt4
If your distribution is not yet using qtchooser, then you need to file
a bug report requesting it and you need to run the full path to qmake.
Note:
- there are some ### left in the buildsystem
Signed-off-by: Thiago Macieira <thiago@macieira.org>
