Re: c++/9872: temporary destructor not called?

[Sandor Kovacs <sandor at ctjapan dot com>]

The following reply was made to PR c++/9872; it has been noted by GNATS.

From: Sandor Kovacs <sandor at ctjapan dot com>
To: Wolfgang Bangerth <bangerth at ticam dot utexas dot edu>
Cc: gcc-bugs at gcc dot gnu dot org,
 <sakovacs at freemail dot hu>,
 <gcc-gnats at gcc dot gnu dot org>
Subject: Re: c++/9872: temporary destructor not called?
Date: Fri, 28 Feb 2003 01:54:43 +0000

 Ok, standalone, no library dependent test code is here: I've basically 
 stripped down the QT QString implementation.
 
 Compile:
 - to eat up memory:
    g++ -c bug.cpp
    g++ bug.o -o bug
 - works properly:
    g++ -fno-elide-constructors -c bug.cpp
    g++ bug.o -o bug
 
 Find .cpp, .ii, .s attached: bug.without-felide.s is the one which eats up 
 memory, the other one bug.with-felide.s is the properly working one.
 
 If this is indeed a bug in gcc then if would explain why I have increasing 
 swap usage after running KDE for weeks, even thoug no app. is running. In 
 such cases I to restart KDE and swap usage disappeares. IF indeed it is a gcc 
 bug and SuSE compiled kde using gcc3.2...
 
 Regards,
 Sandor
 
 > Sandor,
 > I attached your preprocessed source. Thanks for sending them in.
 > However...
 >
 > > If I've missed something or you need further info please let me know;
 >
 > ...yes, we are missing something:
 > >   gcc -L/usr/lib/qt3/lib -lqt-mt bug.o -o bug
 >
 > You are linking with libqt-mt.so. What we need is a _self-contained_
 > testcase, i.e. something that we can test without additional libraries. If
 > we don't have that, it is impossible to find out whether the problem is in
 > the compiler itself, or in the code that is in the library that is being
 > linked in.
 >
 > Please try to pack everything into one file, including the necessary .cpp
 > files from qt, such that it can be compiled without linking together
 > several .o or .so files.
 >
 > Thanks
 >   Wolfgang