On Tue, 2014-11-04 at 03:16 +0100, Tristan Gingold wrote: > On 03/11/14 21:47, Joris van Rantwijk wrote: > > That's great news! I'm glad there will be a release. > > Thanks! > > > There is one patch that I applied in my package which is not yet > > included in the mercurial repository. It is attached to this bug report: > > http://sourceforge.net/p/ghdl-updates/tickets/13/ > > (originally reported by Aníbal Monsalve Salazar) > > > > Without that patch, GHDL fails on MIPS architectures. > > With the patch, it works like a charm on i386, amd64, mips, arm-eabi. > > I have just applied the patch. Somehow I missed the attached patch. > Now applied. >
I believe this patch still needs some work. As it is, if I read it correctly, function type_for_size (ortho_lang.c line 690ff) can potentially return an uninitialised tree in cases where previously it would return NULL_TREE. (Although there may have been other cases - presumably not actually triggered - where it would still return uninitialised data) I don't quite know how restoring that last "else t = NULL_TREE" would interact with the remainder of Joris' patch so I don't know what to do about it. Perhaps initialise t to NULL_TREE at the top of type_for_size? Longer term, there have also been big changes in gcc to support 128 bit types and clean up the size/type mechanism. I can see some study may required to bring ghdl up to date with gcc5.0... but that'll keep for another day. - Brian _______________________________________________ Ghdl-discuss mailing list [email protected] https://mail.gna.org/listinfo/ghdl-discuss
