On Thu, 9 Mar 2006, Dirk Eddelbuettel wrote:
Diethelm,
Any comment from your end?
-- The code only fails on m68k (which has historically had an above-average
share of troubles with Fortran).
-- The previous fSeries released passed so it may be the new Garch code in
file 034B-GarchModelling.f
Yes, from the log
g77 -fPIC -g -O2 -c 034B-GarchModelling.f -o 034B-GarchModelling.o
034B-GarchModelling.f: In subroutine `pc1f01':
In file included from 034B-GarchModelling.f:0:
034B-GarchModelling.f:2628: internal compiler error: in
verify_initial_elim_offsets, at reload1.c:3297
Please submit a full bug report,
It's a compiler bug, and suggests sending a compiler bug report. On such
a minority platform (I last used m68k in the 1980s) and with an
end-of-line gcc version I can see you might not want to do so.
Generally g77 is still more reliable than gfortran at building R packages,
but with 4.0.2 and 4.1.0 the gap has closed a lot. I will probably
migrate to 4.1.0 on Linux/Solaris shortly (MinGW is another story).
-- According to Stephen, gfortran 4.0 compiles it fine so maybe once I turn
from gcc/g77 3.4 to gcc/gfortran 4.0 (and soon 4.1) the problems will go
away by itself. I will also CC Brian as he probably has the best
understanding of the interactions between R and the various gcc releases.
Regards, Dirk
On 9 March 2006 at 13:21, Stephen R Marenka wrote:
| Package: fseries
| Version: 221.10065-1
| Severity: important
|
|
| fseries fails to build from source due to a gcc-3.4 ICE on m68k. For
| what it's worth, I built the package with gfortran-4.0 and it built fine.
|
| I'm not sure of the other implications involved with a change of
| compiler.
|
| You are welcomed to do whatever you want to with this bug report. :)
|
| Here are selected excerpts from the buildd log.
|
|
| | Automatic build of fseries_221.10065-1 on spice by sbuild/m68k 85
| | Build started at 20060221-2154
| |
******************************************************************************
|
| [...]
|
| | ** Using build dependencies supplied by package:
| | Build-Depends: debhelper (>> 4.1.0), r-base-dev (>= 2.0.0), cdbs, r-cran-fbasics
(>= 201.10059), r-cran-fcalendar, r-cran-mgcv (>= 1.1.5), r-cran-vr (>= 7.2.8)
|
| [...]
|
| | Checking correctness of source dependencies...
| | Toolchain package versions: libc6-dev_2.3.5-12
linux-kernel-headers_2.6.13+0rc3-2 gcc-4.0_4.0.2-8 g++-4.0_4.0.2-8
binutils_2.16.1cvs20060117-1 libstdc++6-4.0-dev_4.0.2-8 libstdc++6_4.0.2-8
|
| [...]
|
| | g77 -fPIC -g -O2 -c 034B-GarchModelling.f -o 034B-GarchModelling.o
| | 034B-GarchModelling.f: In subroutine `pc1f01':
| | In file included from 034B-GarchModelling.f:0:
| | 034B-GarchModelling.f:2628: internal compiler error: in
verify_initial_elim_offsets, at reload1.c:3297
| | Please submit a full bug report,
| | with preprocessed source if appropriate.
| | See <URL:http://gcc.gnu.org/bugs.html> for instructions.
| | For Debian GNU/Linux specific bug reporting instructions,
| | see <URL:file:///usr/share/doc/gcc-3.4/README.Bugs>.
| | make[1]: *** [034B-GarchModelling.o] Error 1
| | make[1]: Leaving directory `/build/buildd/fseries-221.10065/src'
| | ERROR: compilation failed for package 'fSeries'
| | ** Removing
'/build/buildd/fseries-221.10065/debian/r-cran-fseries/usr/lib/R/site-library/fSeries'
| | make: *** [R_any_arch] Error 1
| |
******************************************************************************
| | Build finished at 20060221-2216
| | FAILED [dpkg-buildpackage died]
|
|
| A full buildd log is available at
| <http://buildd.debian.org/build.php?pkg=fseries&ver=221.10065-1&arch=m68k>
|
| Other buildd logs may be available at
| <http://buildd.debian.org/build.php?arch=&pkg=fseries>
|
| --
| Stephen R. Marenka If life's not fun, you're not doing it right!
| <[EMAIL PROTECTED]>
|
--
Brian D. Ripley, [EMAIL PROTECTED]
Professor of Applied Statistics, http://www.stats.ox.ac.uk/~ripley/
University of Oxford, Tel: +44 1865 272861 (self)
1 South Parks Road, +44 1865 272866 (PA)
Oxford OX1 3TG, UK Fax: +44 1865 272595
--
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]