Hi,
Continuing my quest to run as many odd and outdated systems as possible, I've now managed to scare up a sparcstation 4 for the home network.
Oddly, given that 3.6-3 runs on i386, ppc, and mac68k versions of NetBSD it crashes reading the sources file on startup with a bus error on SPARC. It happens on the ftell at line 172 of sqFilePluginBasicPrims.c and yet both f and f->file seem to point to reasonable things. This is with gcc 2.9x or so.
Is there something obvious I've missed? The last VM version which does seem to run is 2.8pre3. I'm a bit confused here.
Thanks!
cheers
bruce
The only thing I can think of off hand is that the ftell is trying to resolve into an ftello which is using a long long too allow for > 32bit file pointers. You should confirm if this is happening or not by following how the defines actually resolve into what C code.
On Nov 30, 2004, at 11:14 PM, Bruce ONeel wrote:
Hi,
Continuing my quest to run as many odd and outdated systems as possible, I've now managed to scare up a sparcstation 4 for the home network.
Oddly, given that 3.6-3 runs on i386, ppc, and mac68k versions of NetBSD it crashes reading the sources file on startup with a bus error on SPARC. It happens on the ftell at line 172 of sqFilePluginBasicPrims.c and yet both f and f->file seem to point to reasonable things. This is with gcc 2.9x or so.
Is there something obvious I've missed? The last VM version which does seem to run is 2.8pre3. I'm a bit confused here.
Thanks!
cheers
bruce
-- ======================================================================== === John M. McIntosh johnmci@smalltalkconsulting.com 1-800-477-2659 Corporate Smalltalk Consulting Ltd. http://www.smalltalkconsulting.com ======================================================================== ===
Hi,
Thanks, that was basically it.
The magic change was to change platforms/unix/vm/sqPlatformSpecific.h so that
typedef off_t squeakFileOffsetType
became
typedef long squeakFileOffsetType;
It seems that putting an off_t inside a structure will cause a failure even when you do something like
file->fileSize = 0;
with this compiler/processor/os combo.
That, to be honest, sounds like a bug, but what do I know.
With this small change I'm a happy camper since now all the diverse systems (x86, PPC, SPARC, and m68k) all run 3.6-3 happily.
Thanks again!
cheers
bruce
On Wed, Dec 01, 2004 at 09:45:53AM -0800, John M McIntosh wrote:
The only thing I can think of off hand is that the ftell is trying to resolve into an ftello which is using a long long too allow for > 32bit file pointers. You should confirm if this is happening or not by following how the defines actually resolve into what C code.
On Nov 30, 2004, at 11:14 PM, Bruce ONeel wrote:
Hi,
Continuing my quest to run as many odd and outdated systems as possible, I've now managed to scare up a sparcstation 4 for the home network.
Oddly, given that 3.6-3 runs on i386, ppc, and mac68k versions of NetBSD it crashes reading the sources file on startup with a bus error on SPARC. It happens on the ftell at line 172 of sqFilePluginBasicPrims.c and yet both f and f->file seem to point to reasonable things. This is with gcc 2.9x or so.
Is there something obvious I've missed? The last VM version which does seem to run is 2.8pre3. I'm a bit confused here.
Thanks!
cheers
bruce
--
=== John M. McIntosh johnmci@smalltalkconsulting.com 1-800-477-2659 Corporate Smalltalk Consulting Ltd. http://www.smalltalkconsulting.com ======================================================================== ===
squeak-dev@lists.squeakfoundation.org