Hi, 2009/9/10 Damyan Ivanov <d...@debian.org>: > -=| Damyan Ivanov, Thu, Sep 10, 2009 at 09:47:12AM +0300 |=- >> The patch needs to go into upstream first, so that the >> implementation IDs don't change later when they release sh4 support, >> causing incompatibilities between databases created with Deiban >> package/upstream and upstream/Debian engine. >> >> If you want to take this upstream yourself, you need to send >> a bugreport using their bugtracker[1], attaching a patch to CVS >> HEAD[2] (sending me the link so that I can track it would be nice). If >> not, I'll deal with this somwehere in the future. >> >> [1] http://tracker.firebirdsql.org/ >> [2] http://sourceforge.net/scm/?type=cvs&group_id=9028 > > I have prepared a patch to upstream CVS head. The only missing piece > of information is the alignment rules of sh4. Are there such rules? > For example, may a long long int pointer reside on odd address or not? > > On x86-based hardware unaligned access is ok, but on sparc it is not. > > I am attaching a program that I use to check various hardware details, > like pointer size and such. It is compiled with simple "g++ > archtests.cpp -o archtests". If it runs fine on sh4, then unaligned > access is OK. If you get bus error or similar, unaligned access is > prohibited. >
Thank you. sh4 can not unaligned access is NG in float, double and long double. Does sh4 need define RISC_ALIGNMENT? Bes regards, Nobuhiro -- Nobuhiro Iwamatsu -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org
