214 Commits

Author SHA1 Message Date
zhuk
3b286ff983 Switch to a separate qmake.port.mk. Simplifies logic a lot.
This removes support of separate qmake versions in one port: as we
discovered, there are no ports actually needing this; strangers like
print/poppler don't use qmake in build.

This should be transparent to current ports. But expect more tweaks there:
for now, qt?.port.mk forces qmake.port.mk inclusion, but that will be
reworked to a more common scheme.

Same idea from (at least) espie@ and me; also, espie@ agrees on the plan.
2016-03-10 17:45:11 +00:00
zhuk
7a14b38014 Don't use MAKE_FILE, but hardcode Makefile instead: that's what qmake
actually generates, and some tricky ports like games/gambatte fail when
qmake-specific part tries to use MAKE_FILE variable.
2016-03-09 16:54:28 +00:00
zhuk
ca0f46bce1 Unbreak the case of USE_SYSTRACE=Yes: we tried to run under systrace
always, and bsd.port.mk invokes systrace for us already when in do-build
or do-install target.
2016-03-07 10:13:32 +00:00
zhuk
bac221c6eb Add support for CONFIGURE_STYLE=qmake. No ports affected, since this is
an opt-in. Actual qmake-based ports will be switched on case-by-case
basis in the near future.

Unslacking after discussion with at least naddy@ & espie@ from... 2014?!

Oh my.
2016-03-06 23:59:51 +00:00
zhuk
64c8dd9bd6 Update to Qt 5.5.1.
Some updates, unbreaking stuff, to follow.
2016-03-06 11:56:41 +00:00
zhuk
c104dcee3e Fix fd_set handling bug in Qt5 as well.
okay naddy@
2016-02-22 20:38:08 +00:00
sthen
a0d09ffbad remove devel/flex dep, these are ok with base flex 2015-11-20 11:24:04 +00:00
zhuk
b055d5767f Better later than never: actually update right REVISION-main, and not
the one that went to libastral yesterday.

From Caspar Schutijser, thanks!
2015-10-22 17:26:18 +00:00
zhuk
20cef22fc3 Remove extra REVISION-main.
Pointed out by Caspar Schutijser, thanks!
2015-10-20 21:58:27 +00:00
zhuk
420b39ef2d Allow COPTS/CXXOPTS ha^Wfeature to work in qmake-generated makefiles.
Similar feature existed in qt4 port, but was lost during Qt5 porting.

Inspired by conversation with Caspar Schutijser, should help him on
updating Qt Creator port.
2015-10-20 21:37:26 +00:00
sthen
06e407b972 Qt dlopen()s various libraries including libssl and does runtime
resolution of symbols. It does guard with #ifdefs but since it doesn't
record a hard dependency on the libssl version, an installed package
doesn't automatically get updated to a newer version that notices the
change, so bump qt5's REVISION-main to force an update.

It does fallback sensibly (no crash, just a Warning printf), but since
the change to headers results in a code change to Qt, we need to bump.

Noticed while testing the otter-browser udpate.
2015-09-04 22:08:25 +00:00
ajacoutot
8e5bc8506f Unbreak: libexecinfo header seems to be picked up after the switch to gcc4.9, so
add the corresponding QMAKE_LIBS_EXECINFO to qmake.conf for the time being.
2015-05-29 07:14:44 +00:00
pascal
f0585aa4cf gcc4 bumps, reminded by aja@ 2015-05-28 10:17:22 +00:00
zhuk
66baad7412 Oops, put back backslash that somehow disappeared in r1.37.
Noticed by aja@
2015-05-05 17:43:42 +00:00
ajacoutot
2a0011b8f9 Unbreak: pkg_subst(1) is not in the default path and may not be executable
either.
2015-05-04 21:31:27 +00:00
zhuk
e191b5b217 Improve hack that avoids picking up compiler intrinsics from
/usr/include in favor of /usr/local/lib/gcc ones by Qt5: now it
helps all qmake-qt5-based software as well.

Fixes problems aja@ was seeing with games/gottcode/tanglet at least.
2015-05-04 10:37:01 +00:00
zhuk
48c3d11fc9 Oops, -examples should be bumped, too. 2015-05-03 13:10:59 +00:00
zhuk
4814e31795 Fix a few more glitches in Qt5:
* Add missing Qt5Designer in WANTLIB-examples;
  * Add a gross hack for pkgconfig files being generated to avoid
    value of WRKBUILD mentioned in them;
  * Unhide dependency on geo/geoclue (used by Qt5Location).
2015-05-03 13:00:53 +00:00
zhuk
45ca0232fc It looks like packaging problems naddy@ and aja@ see happen with two files
only. So let's @comment those until the issue is found and fixed.
2015-04-30 21:39:35 +00:00
zhuk
0f44a60a31 It's always better to run update-patches BEFORE commit... 2015-04-25 18:41:30 +00:00
zhuk
c1305c51e5 Remove patch not needed anymore since we have .cpp.o rule in sys.mk now. 2015-04-24 16:31:57 +00:00
zhuk
e7223cf23d Add a deubg printout for qhelpgenerator, which should help to find the root
cause of packaging bug naddy@ observes.

Enable documentation generation again.
2015-04-24 16:16:39 +00:00
zhuk
f4dbdd4ad0 Oops, forgot to remove a few files went to -examples from PLIST-main
during previous commit.
2015-04-24 07:27:34 +00:00
zhuk
3884ab9b31 Sneak in more Qt5 tweaks:
* Replace gross hack named post-configure with a qmake_all target call
    before building documentation. The downside is that rebuild part will
    run in a single process mode; the upside is that qdoc, qhelpgenerator
    and anyone else will work as intended, without need of explicit
    Qt build modules tweaking.

  * Move a few files to -examples. QCounter sensor plugin is commented out
    as it doesn't get installed in case qmake is rerun (see above). It looks
    like this file shouldn't be installed in the first place, but I'm not
    sure, thus using @comment for now.

Documentation building will be disabled until the issue seen by naddy@ gets
fixed.
2015-04-24 07:16:39 +00:00
zhuk
251bb0813d Disable documentation again, bulk builds done by sthen@ and naddy@
continue to reveal dragons here and there. Freinds^WHacking never ends...
2015-04-23 11:49:07 +00:00
zhuk
0613a78bf3 Add another gross hack to make sure qhelpgenerator, when we ask for building
Qt5 documentation, is always picked up from build directory.

Remove IGNORE-qch as things should build fine again (unless a problem naddy@
seen today is unrelated).
2015-04-22 14:12:03 +00:00
zhuk
9e46863387 Drop the patch not needed anymore. Bump REVISION because the affected file
goes into package.
2015-04-22 12:29:34 +00:00
zhuk
a45ebee497 Add IGNORE-qch for now, until the issue with trying to run qhelpgenerator
from LOCALBASE is resolved.

Found by sthen@.
2015-04-20 19:32:38 +00:00
zhuk
7620026661 If file is updated automatically, it should be committed automatically
as well, right?.. Missing distinfo from Qt5 port update to 5.4.1, noticed
by RD Thrush and naddy@.

Sorry. :(
2015-04-20 06:23:13 +00:00
zhuk
9f8f3ac2d5 Update x11/qt5 to 5.4.1. Improvements included:
* Builds fine when another version being installed;
  * Start building documentation, HTML generator bug was fixed (patch is
    waiting for inclusion upstream);

This update do NOT include QtWebEngine: it relies on building from
Chromium sources (bundled with Qt5) and is a real pain to proceed; so lets
concentrate on what we already have.

All dependant ports are checked by building and running.
2015-04-19 13:18:28 +00:00
ajacoutot
aa988913b0 Bump after the recent changes in packaging dbus and dconf. 2015-04-04 09:20:59 +00:00
ajacoutot
d140c22fb9 Fix qtWebkit build with new GLib. 2015-03-27 16:02:32 +00:00
zhuk
0652476dd9 Try to allow Qt5 to get compiled on SPARC: native atomics implementation
was dropped in Qt 5.2, but Qt could use GCC ones - why not?

Initial report from landry@
2015-01-24 22:09:40 +00:00
zhuk
c31af39d7c Avoid last bits of rand() usage in Qt5 libraries. 2015-01-21 20:49:50 +00:00
zhuk
ea1384638b More strcpy() fixes for Qt5. Now, if you see warnings from linker, you
can't reassure yourself that "it's Qt5, not my app" anymore.

Noticed by Adam Wolk (hope now I thank the right person), thanks!
2015-01-21 20:40:46 +00:00
zhuk
83b6a8d8d8 Enable W^X in JIT for Qt5 WebKit.
Tested using Otter Browser on amd64.
2015-01-20 13:38:42 +00:00
zhuk
5d2a0ac57a Make Qt5 pick up certs from /etc/ssl. Fixes annoying warnings in browsers.
Similar (but less invasive) patch was done for Qt4 as well. There is no
valid point for allowing Qt5 to look at many other places, which could
happen be accessible by unauthorized users by occasion, thus we hardcode
the /etc/ssl/ as the only vaild place.
2015-01-20 13:27:59 +00:00
naddy
9b935b0068 <sys/socket.h> already exposes struct iovec, no need to include <sys/uio.h>.
Remove now useless patches.
2015-01-20 11:44:19 +00:00
zhuk
6239cddefd /tmp/cvszOZN00 2015-01-15 11:10:02 +00:00
landry
9784eb5c5e BROKEN-sparc64 = tries to include a non-existent qatomic_sparc.h in src/corelib/thread/qbasicatomic.h 2015-01-12 20:30:18 +00:00
zhuk
cdecce878f Better patch (likely to fix at least some crashes seen in Qt5Webkit),
similar to what real upstream (i.e, WebKitGTK) has.
2015-01-10 22:02:56 +00:00
pascal
6aed10138f Fix in powerpc by linking with --relax.
ok toolchain@ via landry@
2015-01-09 08:10:56 +00:00
zhuk
37882e094d Fix problem with all SQL drivers being probed during CMake-based configure
stage, when not all of them get installed. Actually, it was a matter of
shuffling a few .cmake files between packages.
2015-01-06 21:32:41 +00:00
landry
e148fd3b61 Surprisingly, qt5 fails to build on powerpc, relocation overflow when
linking libQt5WebKit. All hail our glorious binutils!
2015-01-05 18:13:32 +00:00
zhuk
9a16c55c34 Enable W^X in Qt5WebKit/QtDeclarative JIT.
The code looks like initially developed for iOS, but compiles perfectly
on OpenBSD as well. No reason to avoid another hardening that comes
for free, right?
2014-12-26 00:02:32 +00:00
zhuk
361940faf2 On OpenBSD, we override shared library versions. This breaks Qt-GTK
interaction, as found by Adam Wolk (thanks!).

Fix by stopping requiring exact major version of GTK+ libraries.
2014-12-25 23:35:44 +00:00
zhuk
00b4fbef9f My CMake-related expiriments left out Qt5LinguistTools CMake files out
of package; fix that. Should fix the problem aja@ see with new Stellarium.
2014-12-22 21:30:41 +00:00
espie
b14f0f3424 same issue as qt4's qmake. requires nojunk 2014-12-22 18:51:58 +00:00
naddy
c38ae5a9a4 enable parallel building; ok zhuk@ 2014-12-22 18:00:13 +00:00
zhuk
2856c09139 Forgotten to being committed qstrcpy-related patch.
Noticed by naddy@.
2014-12-20 12:15:34 +00:00