Bug#3007: segfault in strcpy is not debuggable
David Engel writes ("Re: Bug#3007: segfault in strcpy is not debuggable"):
> > Package: libc5-dev
> > Version: 5.2.18-6
> > If I write a simple test program which causes a segfault inside
> > strcpy, the stack is corrupted in such a way that gdb is unable to
> > find the correct calling line.
> > I'm reporting this as a bug against libc5-dev as I believe it may be
> > something to do with the compilation options used for the libc;
> > however, it might also be the fault of the compiler, the linker or
> > gdb.
> It has to do with libc.so being compiled with -fomit-frame-pointer.
> You can use the static libc.a if you really need a stack trace.
> Unless you are going to make a case that -fomit-frame-pointer should
> never be used in Debian, I'm going to close this bug report.
Ah, I see. In the days of a.out this was solved by making GCC link
things statically if you linked them with -g. Would it not be
possible to provide a separate libc_g5.so or something which isn't
compiled in this way ?
If there is no better way of solving the problem it ought in any case
to be documented somewhere, IMO ...