Hmmm, that's interesting. The ANTLR2 stuff is part of our source code. We bundle it with the rest of our source here, and it's supposed to be found automatically by cmake.

    https://github.com/sqlitebrowser/sqlitebrowser/tree/master/libs/antlr-2.7.7

Just as a curiosity thing, what happens if you run qmake instead of cmake?

Hi

If I run qmake in the source code directory, nothing happens, qmake exists with status 0.

If I run qmake in the build directory after cmake generated the makefiles, I get an error:

build$ qmake
Usage: qmake [mode] [options] [files]
.....

Doing

$ cmake .
-- Could NOT find Antlr2 (missing:  ANTLR2_LIBRARIES ANTLR2_INCLUDE_DIRS) 

Could it be a path issue?

You have

if(NOT ANTLR2_FOUND)
    set(ANTLR_DIR libs/antlr-2.7.7)
    add_subdirectory(${ANTLR_DIR})
endif()

and this is a relative path, I guess relative to current running directiory and not necessarily to CMAKE_SOURCE_DIR. I always create a build directory outside of the source code and this could be an issue.

Do you usually generate the cmake-Makefile on the same directory as the source code?

Yeah, it could be a path issue.

With cmake, it kind of has a macro like language, so the "if(NOT ANTLR2 ..." clause you're talking about here shouldn't trigger, because it's coming after this line here:

find_package(Antlr2)

That "find_package()" macro runs this code, which seems to be where the problem is:

    https://github.com/sqlitebrowser/sqlitebrowser/blob/master/cmake/FindAntlr2.cmake

The trick will be figuring out what and why. 😉

Btw, with qmake, run it in the main (root) checked out sqlitebrowser git source dir, then run make after it. Like this:

$ git clone https://github.com/sqlitebrowser/sqlitebrowser.git
$ cd sqlitebrowser
$ qmake
$ make
[compiling stuff should happen here]

Note that sometimes when compiling this way on OSX it barfs after a while with an error about missing language translation files. Running make again fixes the problem. (yeah, someone needs to look into that, but it's a pretty minor problem)

Oh man, I'm stupid, I only executed qmake and didn't realize that it only creates the Malefiles.

Now I get

cd libs/antlr-2.7.7/ && make -f Makefile.antlr 
make[1]: Entering directory `/tmp/test/sqlitebrowser/libs/antlr-2.7.7'
make -f Makefile.antlr.Release
make[2]: Entering directory `/tmp/test/sqlitebrowser/libs/antlr-2.7.7'
g++ -c -m64 -pipe -O2 -fPIC -Wall -W -D_REENTRANT -DQT_WEBKIT -DQT_NO_DEBUG -DQT_GUI_LIB -DQT_CORE_LIB -DQT_SHARED -I/usr/share/qt4/mkspecs/linux-g++-64 -I. -I/usr/include/qt4/QtCore -I/usr/include/qt4/QtGui -I/usr/include/qt4 -I. -Irelease -o release/TreeParser.o src/TreeParser.cpp
make[2]: *** [release/TreeParser.o] Segmentation fault
make[2]: Leaving directory `/tmp/test/sqlitebrowser/libs/antlr-2.7.7'
make[1]: *** [release] Error 2
make[1]: Leaving directory `/tmp/test/sqlitebrowser/libs/antlr-2.7.7'
make: *** [sub-libs-antlr-2-7-7-antlr-pro-make_default] Error 2

That's strange

$ g++ -c -m64 -pipe -O2 -fPIC -Wall -W -D_REENTRANT -DQT_WEBKIT -DQT_NO_DEBUG -DQT_GUI_LIB -DQT_CORE_LIB -DQT_SHARED -I/usr/share/qt4/mkspecs/linux-g++-64 -I. -I/usr/include/qt4/QtCore -I/usr/include/qt4/QtGui -I/usr/include/qt4 -I. -Irelease -o release/TreeParser.o src/TreeParser.cpp
Segmentation fault

Last time I had a segfault with GCC it was when I discovered that my RAM was faulty and had to change my ram. That's strange because I have to compile quite a bit of C++ code and I usually don't have these problems.

// edit:

when compiling with cmake, TreeParser.cpp is compiled without a problem.

The trick will be figuring out what and why. 

I'll take a look at that later when I get back home (right now @work). I have some projects using CMake, I have some experience with it.

Sure no worries at all. Wonder if this is somehow showing up a weird gcc bug or something?

I'm not sure, I think I have a problem with my RAM. I'll let memtest run over the weekend.

When I use qmake to compile as root, then the build was complete. But sqlitebrowser crashes immediately with a segfault. If I run it with valgrind (might give me a clue as why it dies with segfault) the QT app window is loaded, rendered but after a couple of seconds it dies with segfault.

As I said, I'll let run memtest over the weekend and see what happens.

Yeah, that does sound kind of weird. 😦

@PSE-AG How did this go?

I didn't forget this issue but I was on holiday and this week has been crazy, lots of deadlines.

I executed memtest over the weekend as I planned, it run more than 60 hour straight without interruptions and not a single error was found. I don't know why gcc fails with segfaults.

I'm still running on a debian 7. I will install debian 8 on virtualbox and see if I can reproduce the same results.

Thanks. Please keep us posted. 😄

@PSE-AG Any luck with this? 😄

Hi

I'm so sorry, I have been terribly busy in the last couple of weeks and I had an operation two weeks ago and have been recovering since then.

I did find out however that the reason I got the segfaults was not my RAM or debian buildchain. Some time ago I installed linuxbrew because I wanted to test something. linuxbrew compiled a glibc, gcc and other packages on its own and put them at the front of PATH. I haven't realized that until I was compiling something for one client of ours and I constantly got segfaults. That's when I executed gcc --version and I got a segfault as well. Then I realized that I was executing the gcc from linuxbrew the whole time.

Since then I haven't had the time to experiment with that. But I haven't forgotten it :)

No worries at all. 😄

Healing well, and getting better is definitely of much higher importance. 😁

So, I make a git clone and the current commit is 684958b29ce3d26684d9142d73df773b11189d54.

I first tried qmake && make and this time it compilied without any issue. The program runs without problems.

I also tried cmake. It still complaining about missing ANTLR2 libraries, but this may be a path issue. cmake is able to generate a Makefile and running make works also as well, the binary runs, everything seems fine.

The only issue I have is that the UI is to aggressive with the focus. If the UI loses focus, it regains automatically again, and this is a problem when you have virtual desktops, because the moment you change to another virtual desktop/screen and the UI loses focus, it regains focus and you jump back at the virtual desktop you've started the UI. I use fluxbox with at 12 virtual screens where every virtual screen has something running. While te UI runs, I'm unable to switch to the other screens.

Cool, making progress. 😄

With the ANTLR2 warnings, there seem to be two distinct paths that can go down:

a) Warning message about it not being found, but DB4S compiles + works ok anyway
b) Warning message, then failure to compile

With a), it can be ignored. It turns out there there are two different approaches for finding ANTLR2 in our CMake configuration, one of them is a fallback to the other. The "ANTLR2 warning but still compiles" situation happens when one of the approaches doesn't work (generating a warning), but the other one does (without giving a "found it!" message 😦 ).

With b) above... that's when it really can't be found. You're not hitting that though, so it sounds like the build step is working "good enough" for you now. 😄

With the focus problem, that's interesting. It's not something I remember anyone mentioning before. It's probably an unexpected interaction from something... but figuring out what exactly is the next challenge. 😄

As a first thought, would you be ok to try compiling against our v3.9.x branch, instead of our master branch?

If the focus problem doesn't happen with the v3.9.x branch, then we'll know it's our recent code which introduced it. That'll give us a much better idea of what to look at. 😄

Hi

the v3.9.x also compiles (qmake/cmake) without a problem. The focus problem is there as well. I'll test it later @home, different system (gentoo).

@PSE-AG How'd this go? 😉

@justinclift: it's me, same person, different account (my personal).

I've made an update in gentoo (dev-db/sqlitebrowser-3.8.0) and it works normal, I can change the to my virtual desktops when sqlitebrowser is running.

I've compiled the v3.9.x version using qmake and cmake and I had the same issue: If I change to another virtual desktop, sqlitebrowser regains immediately the focus and my virtual desktop goes back to that where sqlitebrowser is running.

@shaoran Thanks. 😄

Just to confirm I'm reading that correctly:

• When compiled for the v3.9.x branch, the focus problem is happening
• When compiled for the master branch, the focus problem is not happening

Is that right?

No,

• When compiled for the v3.9.x branch, the focus problem is happening
• When compiled for the master branch, the focus problem is happening
• When compiled for the v3.8.0 branch, the focus problem is not happening

Thanks, now it's clear. It's something we've done between the 3.8.0 and 3.9.0 releases, and it's still happening.

Damn. 😦

We can probably take a brute force approach to working out why. (eg manual bisect)

It's a bit time consuming, although very effective. I won't have time for it personally for at least the next fortnight though. 😦

It looks like you can activate a window using the QWidget::activateWindow function, and we only call it here, but that shouldn't be a problem. So maybe it's an implicit behaviour caused by something else, rather than some explicit call?

Sounds like the focus problem is the same as in #934. But the bi-sected commits mentioned there are from a later date. Maybe it's a connection of multiple things causing this behaviour, I don't know. Either way, if anybody feels like bi-secting this (earlier) problem, please do so 😃

Hi, I can confirm this issue is still present as of commit 425bfa1.

-- Could NOT find Antlr2 (missing: ANTLR2_LIBRARIES ANTLR2_INCLUDE_DIRS)

@EmperorArthur Try running make afterwards, even though that error shows up. There's some indication that the message there is bogus, and might not really mean anything went wrong after all. 😄

Yep, the error message just means that no Antlr2 development files provided by your distribution were found. In that case we just fall back to the Antlr2 library which is included in our repository. So it's safe to just run make as @justinclift said.

However, if you have the Antlr2 lib and dev packages installed and you still see that error message, then that's probably a bug in our cmake file which is worth investigating 😄

The program still compiled, but I spent a bit of time googling around until I found this issue page. If the warning is meant to be ignored, it might be worth it to have the cmake file print out a note.

We've all been trained to treat these sorts of warnings as important. It's really confusing ignoring it. Especially since all the other dependency warnings will cause compilation issues.

I see 😄 I've just changed our cmake script to not print this warning and just silently use the copy of Antlr that's provided in our repository.

@EmperorArthur Would you be ok to try compiling again, using the latest git source, to verify the compile process is warning free for you now? 😄

Confirmed fixed.

#cmake -Wno-dev .
-- The C compiler identification is GNU 5.4.0
-- The CXX compiler identification is GNU 5.4.0
-- Check for working C compiler: /usr/bin/cc
-- Check for working C compiler: /usr/bin/cc -- works
-- Detecting C compiler ABI info
-- Detecting C compiler ABI info - done
-- Detecting C compile features
-- Detecting C compile features - done
-- Check for working CXX compiler: /usr/bin/c++
-- Check for working CXX compiler: /usr/bin/c++ -- works
-- Detecting CXX compiler ABI info
-- Detecting CXX compiler ABI info - done
-- Detecting CXX compile features
-- Detecting CXX compile features - done
-- Configuring done
-- Generating done
-- Build files have been written to: /home/user/src/sqlitebrowser

Awesome! Thanks for checking and letting us know @EmperorArthur. 😄

@PSE-AG The problem with DB4S being too aggressive with it's focus has likely just been fixed in our source code with #1168.

Would you have the time/interest to compile and verify it from your point of view? 😄