Qt5py3 Mac fixes #3790

dakcarto · 2016-11-21T00:31:45Z

Some Mac-specific fixes (and a couple others) for getting QGIS 3 master branch to compile with Homebrew.

@m-kuhn, if you have some time, please review.

Instead of finding Python interpreter, library and framework on Mac, rely upon the reported paths of the interpreter (executable) to decipher whether a framework is being used, then ensure any such framework has its versioned subdirectory Headers used for includes and the base Python library used directly in linking. This removes ambiguity in framework searching, allowing just the PYTHON_EXECUTABLE (user-defined or from FindPythonInterp module) to control which Python is used.

nyalldawson · 2016-11-21T00:50:30Z

src/app/qgsidentifyresultsdialog.cpp

@@ -78,7 +78,9 @@ QgsIdentifyResultsWebView::QgsIdentifyResultsWebView( QWidget *parent ) : QgsWeb
  setSizePolicy( QSizePolicy::MinimumExpanding, QSizePolicy::Minimum );
  page()->setNetworkAccessManager( QgsNetworkAccessManager::instance() );
  // page()->setLinkDelegationPolicy( QWebPage::DelegateAllLinks );
+#ifdef WITH_QTWEBKIT


Is QWebKit not available via homebrew?

Looks like ... qutebrowser/qutebrowser#1357

The Homebrew qt5 formula does have the --with-qtwebkit option, but this requires source building of qt5 and probably pyqt5 as well (though the latter does not have a --with-qtwebkit option, so must opportunistically generate the bindings it QWebKit is available). While this is OK for working with general builds, it would take way too long for Travis CI.

Another option might be to host a qt5-webkit formula that builds against an installed qt5, though I have no idea if the CMake and PyQt5 setups will work right with the module built outside of the Qt5 build chain. Worth trying, as it would be the simplest solution.

Yet another option is to do a PR for Homebrew's qt5 formula to include QWebKit as default now with 5.7+ versions.

QtWebKit is no longer in the main Qt repository, so I guess it's fine to build it in a separate formula.
For reference:
https://github.com/opengisch/osgeo4travis/blob/master/docker/qt5/scripts/build-qtwebkit.sh

m-kuhn · 2016-11-21T15:36:24Z

Hi Larry, I have just read through all the changes and as far as I can tell they look reasonable to me. Unfortunately it didn't build with my self-created qgis3 formula (looks like C++11 is not enabled), so I'm unable to verify the build process.
I would just merge this as-is and fixup possible problems if they appear.

dakcarto · 2016-11-21T18:08:44Z

Hi Larry, I have just read through all the changes and as far as I can tell they look reasonable to me. Unfortunately it didn't build with my self-created qgis3 formula (looks like C++11 is not enabled), so I'm unable to verify the build process.
I would just merge this as-is and fixup possible problems if they appear.

OK, I'll merge tonight and immediately work on the qgis3-dev formula, so as to ensure additional commits are timely.

Not sure when the Travis CI setup can be updated, as we need to solve the QWebKit issue, decide on a newer macOS to build on (10.9 no longer has bottles from Homebrew), and create any ancillary bottles needed for full testing. Also, I'm getting quite a few test fails. Need to check into that, or at a minimum allow failed tests for Mac in the meantime.

m-kuhn · 2016-11-21T18:42:43Z

FYI, We have failing tests with Qt 5.7 on Linux as well, see #3400 .

Concerning PyQtWebKit, the Qt 4.8 version of these bindings are built as part of QGIS 2.18 on some platforms, you might have a look at that: https://github.com/qgis/QGIS/tree/release-2_18/python/QtWebKit

edit: I don't think we should have that in QGIS 3.0 but this might help to build a separate formula.

dakcarto · 2016-11-23T04:41:15Z

@m-kuhn, @nyalldawson, I've added a separate formula for QtWebKit/QtWebKitWidgets.

That formula builds OK, but it seems to have an issue with /usr/local/Cellar/qt5/5.7.0/plugins/bearer/libqcorewlanbearer.dylib. If I pull that plugin out, the test passes just fine. I am trying a QGIS master build against this setup. If the problem persists at QGIS runtime, I may just rebuild and bottle via brew install qt5 --with-webkit and see if that helps.

dakcarto · 2016-11-23T04:57:58Z

Hmm. QtWebKit (by way of separate formula) seems to work fine in QGIS. So, I'll try to complete a qgis3-dev formula ASAP.

nyalldawson · 2016-11-23T05:42:07Z

@dakcarto I'm getting:

In file included from /usr/include/qt5/QtCore/qchar.h:43:0,
                 from /usr/include/qt5/QtCore/qstring.h:48,
                 from /usr/include/qt5/QtCore/QString:1,
                 from /usr/include/qt5/QtCrypto/qca_core.h:36,
                 from /usr/include/qt5/QtCrypto/qca.h:36,
                 from /usr/include/qt5/QtCrypto/QtCrypto:1,
                 from qcaossl.cpp:2:
/usr/include/qt5/QtCore/qglobal.h:1113:4: error: #error "You must build your code with position independent code if Qt was built with -reduce-relocations. " "Compile your code with -fPIC (-fPIE is not enough)."
 #  error "You must build your code with position independent code if Qt was built with -reduce-relocations. "\

on master. Any ideas?

nirvn · 2016-11-23T07:55:22Z

Same here, can't build master ATM.

m-kuhn · 2016-11-23T08:33:36Z

I have partially reverted this for now in 9bb3235 (basically the one line that prevented it from even checking for the qca_ossl plugin)

It's a bit puzzling that -DCMAKE_POSITION_INDEPENDENT_CODE=ON results in -fPIE and not -fPIC. Is this a compiler-dependent result?

m-kuhn · 2016-11-23T10:26:21Z

https://stackoverflow.com/questions/40761777/how-to-pass-fpic-in-a-try-run-command

dakcarto added 9 commits November 20, 2016 15:30

Use Qt5 non-OpenSSL define in testing for SSL support

4c2725b

Check for Qt5 subdirectory in PYQT5_SIP_DIR

e9fa3e0

Check for Qt5-specific version of Mac qca framework

74dfd1f

Update finding QCA OSSL plugin for Qt5 and C++11

9bb3235

[auth] Remove dependency on SSL cert utils from auth config class

a510516

Skip missing functionality in identity dialog if QWebKit not installed

64b6a44

Add a CMake debug include, with simple variables dump for development

c123d3b

Add CMAKE_POSITION_INDEPENDENT_CODE flag to QCA test for OSSL plugin

eca83e3

nyalldawson reviewed Nov 21, 2016

View reviewed changes

dakcarto merged commit d712b49 into qgis:master Nov 23, 2016

qgib mentioned this pull request May 24, 2019

OTF Raster Reprojection artifacts (at some scales) #13848

Closed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Qt5py3 Mac fixes #3790

Qt5py3 Mac fixes #3790

dakcarto commented Nov 21, 2016

nyalldawson Nov 21, 2016

m-kuhn Nov 21, 2016

dakcarto Nov 21, 2016

m-kuhn Nov 21, 2016

m-kuhn commented Nov 21, 2016

dakcarto commented Nov 21, 2016

m-kuhn commented Nov 21, 2016 •

edited

Loading

dakcarto commented Nov 23, 2016

dakcarto commented Nov 23, 2016

nyalldawson commented Nov 23, 2016

nirvn commented Nov 23, 2016

m-kuhn commented Nov 23, 2016

m-kuhn commented Nov 23, 2016

Qt5py3 Mac fixes #3790

Qt5py3 Mac fixes #3790

Conversation

dakcarto commented Nov 21, 2016

nyalldawson Nov 21, 2016

Choose a reason for hiding this comment

m-kuhn Nov 21, 2016

Choose a reason for hiding this comment

dakcarto Nov 21, 2016

Choose a reason for hiding this comment

m-kuhn Nov 21, 2016

Choose a reason for hiding this comment

m-kuhn commented Nov 21, 2016

dakcarto commented Nov 21, 2016

m-kuhn commented Nov 21, 2016 • edited Loading

dakcarto commented Nov 23, 2016

dakcarto commented Nov 23, 2016

nyalldawson commented Nov 23, 2016

nirvn commented Nov 23, 2016

m-kuhn commented Nov 23, 2016

m-kuhn commented Nov 23, 2016

m-kuhn commented Nov 21, 2016 •

edited

Loading