Bug #48751 for AI-MegaHAL: BYTE4 defined as unsigned long breaks on 64 bit platforms

Sun Aug 16 08:44:32 2009 harv [...] ruin.nu - Ticket created

Subject:

BYTE4 defined as unsigned long breaks on 64 bit platforms

In libmegahal.c, on line 154 in 0.07 (146 in 0.06) there is: #define BYTE4 unsigned long However, unsigned long is 8 bytes on many 64 bit platforms, so loading a brain generated on a 32 bit platform will not work. However, the fix is easy, just changing it to: #define BYTE4 unsigned int That might break on older platforms, where int is 16 bit, not sure how common those are these days though.

Sun Oct 18 23:05:14 2009 https://www.google.com/accounts/o8/id?id=AItOawnlNzkbyGCzSmAN8oBQ4OFZYqObWyCwc1I - Correspondence added

From:

Dave

Attached a patch for this bug. The only place it seemed to cause a problem was in load_dictionary(), as fread() was reading an unsigned long into an int. Just changed the int to an unsigned long. On Sun Aug 16 08:44:32 2009, harv wrote: Show quoted text

> In libmegahal.c, on line 154 in 0.07 (146 in 0.06) there is: > > #define BYTE4 unsigned long > > However, unsigned long is 8 bytes on many 64 bit platforms, so loading > a brain generated on a 32 bit platform will not work. However, the fix > is easy, just changing it to: > > #define BYTE4 unsigned int > > That might break on older platforms, where int is 16 bit, not sure how > common those are these days though.

Download 64bit-patch
application/octet-stream 522b

Message body not shown because it is not plain text.

Sun Oct 18 23:05:18 2009 The RT System itself - Status changed from 'new' to 'open'

Mon Oct 19 06:27:26 2009 harv [...] ruin.nu - Correspondence added

On Sun Oct 18 23:05:14 2009, https://www.google.com/accounts/o8/id? id=AItOawnlNzkbyGCzSmAN8oBQ4OFZYqObWyCwc1I wrote: Show quoted text

> Attached a patch for this bug. The only place it seemed to cause a > problem was in load_dictionary(), as fread() was reading an unsigned > long into an int. > > Just changed the int to an unsigned long.

That's still a problem, since unsigned long is still 8 bytes, and not 4, it'll read too much from the file. With this change I'm getting a segmentation fault instead of an unusable brain.