Ben Collins <[EMAIL PROTECTED]> writes: > On Mon, Apr 23, 2001 at 09:37:09AM -0700, Philippe Troin wrote: > > Ben Collins <[EMAIL PROTECTED]> writes: > > > > > On Sun, Apr 22, 2001 at 11:31:18PM -0700, Philippe Troin wrote: > > > > [EMAIL PROTECTED]:~% gcc -o foo foo.cpp -lstdc++ > > > > > > IMO, you should use: > > > > > > % g++ foo foo.cpp > > > > This brings the same result as 'gcc -o foo foo.cpp -lstdc++': crash. > > > > > ...instead. There are a lot more things going on when using g++ other > > > than adding -lstdc++. > > > > According to 'gcc -v' the only difference between `g++ -o foo foo.cpp' > > and `gcc -o foo foo.cpp' is the two extra libraries `-lstdc++ -lm' at > > link time... > > > > > If this works for you, then I suggest closing the bug. > > > > If it were that easy... > > > > As mentionned in the initial bug report, I found this with a much > > bigger program that was crashing when throwing complex (instances of > > derived classes) exceptions... This program was initially compiled and > > linked with g++. This program does not need libstdc++. When I tried to > > link it without (with gcc), it started working. > > > > Just adding '-lstdc++' makes exceptions handling buggy... The crash > > happens during the throw()... I'm quite puzzled... > > Interesting. It works perfectly fine for me under sparc (even with your > command line). Of course, I am using a gcc-3.0 snapshot more recent than > that in the archive. I'll be uploading a new snapshot within a day or > two, so please give that a try.
The aleph package (this is the bigger c++ program that led me to discover this) has problems with this (works linked without stdc++, breaks with stdc++ linked in) only on arm and mips. The much simplier test program only fails on mips, but I would suspect it's the same underlying cause. Phil.
