On 12/01/2009 12:38 AM, Erik van Pienbroek wrote:
Now that Fedora 13 development is opened this is the perfect moment
switch to DWARF2. I'm not too familiar with compiler internals, but I
don't see any reason why we shouldn't perform this switch. Being more
compatible with MSVC is a big plus for MinGW in general. Getting rid of
libgcc_s_sjlj-1.dll is also a plus for me. So I would say: go for it.
I don't think it's possible to have cross-dll exceptions without a
shared helper DLL. In any case, it's true that we'd get rid of
libgcc_s_sjlj-1.dll, but it would be replaced by libgcc_s_dw2-1.dll.
If things turn out to be broken, we can always revert to the old
exception model as we still have until February before the Fedora 13
feature freeze .
I'm very interested in having a stable mingw toolchain in Fedora. This
is the reason why I started this work now; as opposed to waiting for
someone else to step up and propose the patches at a later time.
As you say, now it's the perfect moment to do the switch. If we decide
to switch over, we should do it NOW, so that we'd have enough time to
shake out remaining bugs or even revert back to SJLJ.
You mentioned a list of packages which currently depend on
libgcc_s_sjlj-1.dll. Do you think it will be sufficient to just rebuild
those packages or is a (coordinated) mass-rebuild of all mingw packages
Most of our mingw32- packages are plain C code, which don't use
exceptions. The change only affects C++ programs, and shouldn't have any
effect for anything else. However, to be on safe side, _if_ we are going
to switch over, it'd be better to do it before F-13 mass rebuild.
Does this change have any effect on gdb compatibility?
I think not, but I am not very sure. We don't have mingw32-gdb in Fedora
and trying with a gdb binary provided by upstream mingw probably doesn't
show anything meaningful.