[Bug middle-end/15937] [3.3 only] miscompiles gengtype of HEAD as bootstrap compiler

rguenth at tat dot physik dot uni-tuebingen dot de gcc-bugzilla@gcc.gnu.org
Thu Jun 17 07:48:00 GMT 2004


------- Additional Comments From rguenth at tat dot physik dot uni-tuebingen dot de  2004-06-17 07:48 -------
Subject: Re:  [3.3 only] miscompiles gengtype of HEAD
 as bootstrap compiler

On Thu, 17 Jun 2004, wilson at gcc dot gnu dot org wrote:

>
> ------- Additional Comments From wilson at gcc dot gnu dot org  2004-06-17 06:23 -------
> I am not able to reproduce this problem.
>
> I tried first on a debian testing system.  I bootstrapped the gcc-3.3.4 tarball
> with --enable-languages=c, and then tried to bootstrap mainline.  gengtype
> worked correctly.
>
> I then tried on a debian unstable system.  I ran apt-get to make sure everything
> was up to date, then tried to bootstrap mainline using /usr/bin/gcc which is
> gcc-3.3.4-1.  gengtype worked correctly.
>
> I suspect that this may be a bison/flex versioning problem rather than a gcc
> optimizer bug.  Perhaps the debian folks have already updated the bison/flex
> packages to make this problem go away, or perhaps you have out of date versions
> of the bison/flex packages, or perhaps someone has already applied a patch to
> the bison/flex input files on mainline that fixed the problem.  There is no easy
> way for me to tell.

I'll try again soon.

> By the way, reporting a bug against mainline is a bit ambiguous, unless you
> indicate when you checked out mainline.  If you use contrib/gcc_update, then you
> can include the LAST_UPDATED file.  Otherwise, you can try attaching the top
> entry of the relevant ChangeLog file, e.g. gcc/ChangeLog for gcc problems.

Yes, sorry.  The day of the bugreport, of course.

> I notice that gengtype was compiled without optimization.  If you configure
> normally, it will be compiled with -O2.  That implies that you did something
> different when configuring that you failed to specify in the bug report.  This
> could perhaps explain why I could not reproduce the problem.

Really?  I expected (and experienced) the host compiler to be used only
with -O0 for compiling stage1.

Richard.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=15937



More information about the Gcc-bugs mailing list