Mail Archives: pgcc/1999/08/05/15:24:41
Marc Lehmann wrote:
>
> On Sun, Aug 01, 1999 at 08:32:58PM +0200, jkf AT casema DOT net wrote:
> > > It's also in CVS (pgcc-2_95).
> > >
> > > Enjoy.
> > >
> >
> > I tried the patch
>
> I will need to re-do the patch. However, due to time and
> bandwidth-limitations I can't do a diff against the gcc-2.95.tar.gz, only
> against the cvs version (which has subtle differences).
>
> And Bernhard, who is usually making (and checking!) the patches is
> currently away, so we might wait until 2.95.1 (which should be out in
> 1999-08-15).
>
> > What seems more serious, using the same pgcc CVS compiler,
> > pgcc-2.95 crashes, and gcc-2.95 goes on.
>
> (how did it crach when you couldn't built it?)
>
> > Here are the final reports from building pgcc-2.95:
>
> I can't explain these. But:
>
> > gcc -DIN_GCC -march=pentiumpro -O6 -o ../jc1 \
>
> These are not the standard flags, which suggests you dsidn't follow the
> standard installation instructions. Following these and doing a bootstrap
> would most likely fix these problems.
>
Ok, I might not have been too clear
I have installed a working pgcc (correctly working)
of last saturday from CVS, when copying the entire
source tree (I never make builds using this tree directly,
to ensure it remains clean)
When I do this
cd /tmp
mkdir pgcc-cvs
cd pgcc-cvs
cp -a $HOME/pgcc-in/pgcc .
mkdir pgcc-build
cd pgcc-build
../pgcc/configure --prefix=/usr
make
this works, and gives a working pgcc
when I do this
cd /tmp
mkdir pgcc-2.95
cd pgcc-2.95
bunzip2 < $HOME/linupdates/gcc-2.95.tar.bz2 | tar -xv
cd gcc-2.95
gunzip < $HOME/linupdates/gcc-2.95-pgcc-2.95.diff.gz | patch -p1
cd ..
mkdir pgcc-build
cd pgcc-build
../gcc-2.95/configure --prefix=/usr
make
the building stops with the messages in my previous mail.
When i do this
cd /tmp
mkdir gcc-2.95
cd gcc-2.95
bunzip2 < $HOME/linupdates/gcc-2.95.tar.bz2 | tar -xv
mkdir gcc-build
cd gcc-build
../gcc-2.95/configure --prefix=/usr
make
this gives a working gcc-2.95
(I just verified this again on a fresh rebooted system,
with three equal consoles started from xwindows)
the -march=pentiumpro -O6 got in via setting the
CFLAGS and CXXFLAGSenvironment variables
a build using make bootstrap also fails on the
patched version, not on the other two.
- Raw text -