Mail Archives: djgpp-workers/2004/04/15/11:12:20
On Tuesday 06 April 2004 18:06, Andris Pavenis wrote:
> Test archives of 2004/04/01 CVS version of gcc-3.4.0 prerelease for DJGPP
> are available at
>
> ftp://ftp.delorie.com/pub/djgpp/alpha/v2gnu/
>
> No real testing done (except of bootstrapping it, so at least it doesn't
> crash while bootstrapping).
>
> One thing which must be tested is precompiled headers. I disabled use of
> them for building libstdc++, as earlier I had problems with them, but this
> not tested already for more than a month.
Did some tests with it while was in Latvia for Easter time
Problems
========
1) /dev/env/DJDIR/include instead of DOS style directory for include
file dependencies.
2) ICE (segmentation fault) building one file from glibmm-2.3.8. Similar
problem in one of my sources (running gcc in a different way workarounds
the problem). Also ICE for 4 files, when building cln-1.1.6.
3) Had problems in my sources with declaring binary operators << and *
involving my template class as friends of this class. Rewrote code
without friend declarations (member function instead for <<, static
member function which performs *, leaving binary operators as
inline functions which only calls member functions.
Packages built (tests under Windows XP)
=======================================
C:
==
gdb-6.1 builds without problems. Seems to work Ok. Had to modify
bfdsymify.c to build it. No changes needed for GDB itself.
glib-2.4.0 builds without problems. The number of test failures (5)
is the same as with gcc-3.3.3. Also patches which I applied
are the same.
libxml2-2.6.8 Builds without problems (some patches needed, same as for
gcc-3.3.3)
C++:
====
libsigc++-1.9.16 builds without problems, all tests passes
glibmm-2.3.8 ICE (segmentation fault) compiling glib/glibmm/shell.cc
when make invoked not directly in that directory, but
from top level or glib/. If running make
in glib/glibmm/shell.cc, then there is no ICE.
Other stuff builds Ok
libxml++-2.5.2 builds without problems. Test examples seem to work OK.
cln-1.1.6 4 ICE (segfaults in GCC). Can be workarounded by changing
order of include files. After that all included tests
passes.
GiNaC-1.2.0 Builds OK. Some test failures, partialy due to use of
cmp for comparisson of text files. Not all tests performed,
as some of them were terribly slow.
Andris
- Raw text -