Le jeu 16/01/2003 à 22:39, Matthias Klose a écrit : > Josselin Mouette writes: > > Package: gcc-3.2 > > Version: 1:3.2.2-0pre3 > > Severity: important > > > > It seems that the m68k buildd encountered a GCC bug while building HDF5 > > with gcc-3.2 : > > http://buildd.debian.org/fetch.php?&pkg=hdf5&ver=1.4.4-1&arch=m68k&stamp=1042664726&file=log&as=raw > > The appropriate lines are : > > gcc -Wno-long-long -ansi -pedantic -Wall -W -Wundef -Wshadow > > -Wpointer-arith -Wbad-function-cast -Wcast-qual -Wcast-align > > -Wwrite-strings -Wconversion -Wsign-compare -Waggregate-return > > -Wstrict-prototypes -Wmissing-prototypes -Wmissing-declarations > > -Wredundant-decls -Wnested-externs -Winline -O -I. -I../../../src > > -DH5_DEBUG_API -DNDEBUG -D_LARGEFILE_SOURCE -D_LARGEFILE64_SOURCE > > -D_FILE_OFFSET_BITS=64 -D_POSIX_SOURCE -D_BSD_SOURCE -c > > ../../../src/H5Fistore.c -fPIC -DPIC -o .libs/H5Fistore.lo > > ../../../src/H5Fistore.c: In function `H5F_istore_read': > > ../../../src/H5Fistore.c:1777: Internal compiler error in > > instantiate_virtual_regs_1, at function.c:4134 > > > > I don't have much time to investigate on it yet, but I will try to > > extract the faulty code in a few days. > > yes, _PLEASE_ > > - attach the preprocessed source
The preprocessed source is quite huge (~200K), even when removing all the other functions, as there are lots of required headers. It is available on gluck:~joss/public_html/foo.i. > - name the gcc versions that built your package It looks like GCC 2.95.4 and GCC 3.0.4 don't exhibit that bug. > - investigate, if changing the optimization level / omitting -f > options changes the behaviour. The only faulty seems to be -O1. With -O0 or -O2, the file builds fine (which makes me wonder if this package shouldn't be built with -O2 anyway). Regards, -- .''`. Josselin Mouette /\./\ : :' : [EMAIL PROTECTED] `. `' [EMAIL PROTECTED] `- Debian GNU/Linux -- The power of freedom
signature.asc
Description: PGP signature
