Qt-interest Archive, August 2006
Removing mingwm10.dll dependency
Message 1 in thread
Dear List,
I am using open source Qt4.1.4 and to build a Windows binary I use MinGW (or
MingW, whatever it is). I was frustrated by the mingwm10.dll dependency, so
I searched around and found this posting that was helpful:
http://lists.trolltech.com/qt-interest/2005-10/thread00381-0.html#msg00553
The problem is that I have to manually remove the -mthreads option from the
LFLAGS = line in my Makefile. It works but I'd like to tell qmake to just
not put -mthreads in there in the first place. I tried adding
LFLAGS -= -mthreads
to my .pro file, but it didn't accomplish anything. Any tips on how I can
avoid this by-hand edit every time I compile? Thanks in advance.
(Note that Makefile doesn't actually contain a reference to -mthreads, it's
Makefile.Debug and Makefile.Release that do.)
Nathan
--
[ signature omitted ]
Message 2 in thread
Nathan Carter wrote:
I tried adding
>
> LFLAGS -= -mthreads
>
Have you tried it with LIBS -= -mthreads ?
--
[ signature omitted ]
Message 3 in thread
Hi,
> I tried adding
>>
>> LFLAGS -= -mthreads
>>
Try QMAKE_LFLAGS or QMAKE_LFLAGS_THREAD instead:
http://doc.trolltech.com/4.1/qmake-variable-reference.html#qmake-lflags-thread
> Have you tried it with LIBS -= -mthreads ?
The -mthreads option does not specify a library. I'm afraid this won't
work.
Anyway, Qt requires thread support, so I don't think removing -mthreads
can work.
--
[ signature omitted ]
Message 4 in thread
Hi,
> Anyway, Qt requires thread support, so I don't think removing -mthreads
> can work.
NOT true, because "-mthreads" is just needed for thread-safe exception
handling. Since Qt DOES NOT use exceptions at all, you can safly remove
this flag on a MinGW32-System.
Here are the steps I did to get it to work without dependency:
1.) edit [QT]\mkspecs\win32-g++\qmake.conf: remove all occurrences of
"-mthreads".
2.) Recompile Qt
3.) Recompile your project -> dependency will no longer exist!
I mentioned this to Trolltech Support twice, they confirmed the fact, but
nothing happened ever since. I have to do this hack everytime I install a
new Qt version :-(
Regards and good luck,
Malte
--
[ signature omitted ]
Message 5 in thread
Thank you very much! I appreciate this fix; it helps a lot, since I don't
care about C++ exceptions in my project.
Nathan
>> Anyway, Qt requires thread support, so I don't think removing -mthreads
>> can work.
>
> NOT true, because "-mthreads" is just needed for thread-safe exception
> handling. Since Qt DOES NOT use exceptions at all, you can safly remove
> this flag on a MinGW32-System.
>
> Here are the steps I did to get it to work without dependency:
>
> 1.) edit [QT]\mkspecs\win32-g++\qmake.conf: remove all occurrences of
> "-mthreads".
>
> 2.) Recompile Qt
>
> 3.) Recompile your project -> dependency will no longer exist!
--
[ signature omitted ]
Message 6 in thread
Hi,
>> Anyway, Qt requires thread support, so I don't think removing -mthreads
>> can work.
>
> NOT true, because "-mthreads" is just needed for thread-safe exception
> handling. Since Qt DOES NOT use exceptions at all, you can safly remove
> this flag on a MinGW32-System.
Qt actually may throw a bad_alloc exception in out-of-memory situations.
Client slots connected to Qt signals may also throw exceptions. And this
may happen in different threads.
The -mthreads documentation reads:
http://gcc.gnu.org/onlinedocs/gcc-4.1.1/gcc/i386-and-x86_002d64-Options.html#index-maccumulate_002doutgoing_002dargs-1041
Support thread-safe exception handling on `Mingw32'. Code that
relies on thread-safe exception handling must compile and link
all code with the -mthreads option.
I'm not sure what "thread-safe exception handling" exactly means and
what the real scope of "compile and link all code with the -mthreads
option" is. It may be lack of knowledge and testing on my side, but I'm
not convinced so far that -mthreads can be removed safely.
--
[ signature omitted ]
Message 7 in thread
Hi Dimitri,
I had a discussion with your support stuff about this issue before. The
result was that I could safely remove the flag. I compiled with
"-no-exceptions" flag though...
I am working with "-mthreads"-less Qt library for ages now and never had
any regarding issues or problems. I am pretty sure that you CAN safely
remove the flag. I am not sure if it is important to issue
"-no-exceptions" flag additionally on compile-time of Qt.
BTW, I refer to Qt 4.x. - I don't know about 3.x.
Regards,
Malte
Dimitri <dimitri@xxxxxxxxxxxxx> schrieb am 30.08.2006 22:55:55:
> Hi,
>
> >> Anyway, Qt requires thread support, so I don't think removing
-mthreads
> >> can work.
> >
> > NOT true, because "-mthreads" is just needed for thread-safe exception
> > handling. Since Qt DOES NOT use exceptions at all, you can safly
remove
> > this flag on a MinGW32-System.
>
> Qt actually may throw a bad_alloc exception in out-of-memory situations.
> Client slots connected to Qt signals may also throw exceptions. And this
> may happen in different threads.
>
> The -mthreads documentation reads:
> http://gcc.gnu.org/onlinedocs/gcc-4.1.1/gcc/i386-and-x86_002d64-
> Options.html#index-maccumulate_002doutgoing_002dargs-1041
> Support thread-safe exception handling on `Mingw32'. Code that
> relies on thread-safe exception handling must compile and link
> all code with the -mthreads option.
>
> I'm not sure what "thread-safe exception handling" exactly means and
> what the real scope of "compile and link all code with the -mthreads
> option" is. It may be lack of knowledge and testing on my side, but I'm
> not convinced so far that -mthreads can be removed safely.
>
> --
>
> Dimitri
>
> --
>
> To unsubscribe - send a mail to qt-interest-request@xxxxxxxxxxxxx
> with "unsubscribe" in the subject or the body.
> List archive and information: http://lists.trolltech.com/qt-interest/
>
--
[ signature omitted ]
Message 8 in thread
Hi,
> I had a discussion with your support stuff about this issue before. The
> result was that I could safely remove the flag. I compiled with
> "-no-exceptions" flag though...
If you compile with the -fno-exceptions flag, exceptions will not
propagate through Qt. Any such attempt will crash the program.
> I am working with "-mthreads"-less Qt library for ages now and never had
> any regarding issues or problems. I am pretty sure that you CAN safely
> remove the flag. I am not sure if it is important to issue
> "-no-exceptions" flag additionally on compile-time of Qt.
This proves *you* can safely remove the flag in a specific context,
because no exceptions are thrown on out-of-memory conditions and
probably never propagate exceptions through Qt. This doesn't prove
*everyone* can safely remove the flag.
Now it may be that it's useless for most Qt users, and the default
should therefore be to avoid it. Instead of removing -mthreads from
mkspecs/win32-g++/qmake.conf altogether, I would recommend a choice between:
* supporting exceptions: use -mthreads
* not supporting exceptions: use -fno-exceptions
This choice would be handled by the Windows installer. Since I don't
master both the Windows installer and the Unix configure script, I can't
modify this myself, but you could suggest this fix to the Trolltech support.
In any case, before fixing anything, I think we need to clarify in the
compiler documentation what "thread-safe exception handling" exactly
means and what the real scope of "compile and link all code with the
-mthreads option" is.
--
[ signature omitted ]
Message 9 in thread
> If you compile with the -fno-exceptions flag, exceptions will not
> propagate through Qt. Any such attempt will crash the program.
Ack'ed! But this is desired in my case.
>
> > I am working with "-mthreads"-less Qt library for ages now and never
had
> > any regarding issues or problems. I am pretty sure that you CAN safely
> > remove the flag. I am not sure if it is important to issue
> > "-no-exceptions" flag additionally on compile-time of Qt.
>
> This proves *you* can safely remove the flag in a specific context,
> because no exceptions are thrown on out-of-memory conditions and
> probably never propagate exceptions through Qt. This doesn't prove
> *everyone* can safely remove the flag.
True as well - but I bet that there are many users out there having
similar side conditions.
> Now it may be that it's useless for most Qt users, and the default
> should therefore be to avoid it. Instead of removing -mthreads from
> mkspecs/win32-g++/qmake.conf altogether, I would recommend a choice
between:
> * supporting exceptions: use -mthreads
> * not supporting exceptions: use -fno-exceptions
> This choice would be handled by the Windows installer. Since I don't
> master both the Windows installer and the Unix configure script, I can't
> modify this myself, but you could suggest this fix to the Trolltech
support.
Sounds like a good approach to me :)
> In any case, before fixing anything, I think we need to clarify in the
> compiler documentation what "thread-safe exception handling" exactly
> means and what the real scope of "compile and link all code with the
> -mthreads option" is.
True!
Regards and thanks for your considerations,
Malte
--
[ signature omitted ]
