Skip Menu |
 

History Show all quoted text — Show full headers
# Mon Jul 07 15:08:09 2008 lxs (Alexandra Ellwood) - Ticket created
From: lxs@mit.edu
Subject: SVN Commit
Download (untitled) / with headers
text/plain 180B

Since we are copying from one structure to another, copy elements.
Using memcpy is fragile.

Commit By: lxs



Revision: 20499
Changed Files:
U trunk/src/lib/krb5/krb/gic_opt.c
# Mon Jul 07 15:08:12 2008 lxs (Alexandra Ellwood) - Requestor lxs (Alexandra Ellwood) added
# Mon Jul 07 15:08:12 2008 lxs (Alexandra Ellwood) - Status changed from 'new' to 'open'
# Mon Jul 07 15:17:09 2008 lxs (Alexandra Ellwood) - Component krb5-libs added
# Mon Jul 07 15:17:09 2008 lxs (Alexandra Ellwood) - Version_reported 1.6.5 added
# Mon Jul 07 15:17:10 2008 lxs (Alexandra Ellwood) - Target_Version 1.7 added
# Mon Jul 07 15:17:54 2008 lxs (Alexandra Ellwood) - Given to tlyu (Taylor Yu)
# Mon Jul 07 15:17:54 2008 lxs (Alexandra Ellwood) - Comments added
Download (untitled) / with headers
text/plain 32B
Assigned to Tom for code review.
# Tue Jul 08 19:40:14 2008 tlyu (Taylor Yu) - Correspondence added
To: rt@krbdev.mit.edu
Subject: Re: [krbdev.mit.edu #6010] krb5int_gic_opte_copy should copy elements individually
From: Tom Yu <tlyu@MIT.EDU>
Date: Tue, 08 Jul 2008 19:40:03 -0400
RT-Send-Cc:
Download (untitled) / with headers
text/plain 439B
On most platforms, the memcpy() should work because the structure
layouts are identical for their initial overlapping elements. (The C
standard does not guarantee this.) On the Mac, the use of "#pragma
pack" could cause a difference in structure layout.

We need to investigate whether this constitutes an ABI change on the
Mac. Given that a caller can receive a new structure type-punned to
the old one, I believe it is an ABI change.
# Wed Jul 09 12:41:39 2008 raeburn (Ken Raeburn) - Comments added
From: Ken Raeburn <raeburn@MIT.EDU>
To: rt-comment@krbdev.mit.edu
Subject: Re: [krbdev.mit.edu #6010] krb5int_gic_opte_copy should copy elements individually
Date: Wed, 9 Jul 2008 12:41:28 -0400
RT-Send-Cc:
Download (untitled) / with headers
text/plain 1008B
On Jul 8, 2008, at 19:40, Tom Yu via RT wrote:
Show quoted text
> On most platforms, the memcpy() should work because the structure
> layouts are identical for their initial overlapping elements. (The C
> standard does not guarantee this.) On the Mac, the use of "#pragma
> pack" could cause a difference in structure layout.
>
> We need to investigate whether this constitutes an ABI change on the
> Mac. Given that a caller can receive a new structure type-punned to
> the old one, I believe it is an ABI change.

Yes, I believe it is as well. I think the only way to avoid that is a
Mac-specific "#pragma pack" for the extended version of the structure
too. It will have no effect on other platforms, and the only Mac code
it could break would be code trying to explicitly set one field after
the extended structure has been put into use, and somehow relying on
the fact that it *doesn't* set the intended field, but overwrites
something else. I won't lose any sleep over breaking such programs. :)

Ken
# Thu Jul 10 18:34:30 2008 tlyu (Taylor Yu) - Correspondence added
From: tlyu@mit.edu
Subject: SVN Commit
Download (untitled) / with headers
text/plain 210B

Use #pragma pop on the Mac to ensure that the krb5_gic_opt_ext
structure has the same layout as the public version.

Commit By: tlyu



Revision: 20511
Changed Files:
_U trunk/
U trunk/src/include/k5-int.h
# Thu Jul 10 18:34:44 2008 tlyu (Taylor Yu) - Correspondence added
From: tlyu@mit.edu
Subject: SVN Commit
Download (untitled) / with headers
text/plain 146B

Add comment and <TargetConditionals.h> to previous.

Commit By: tlyu



Revision: 20512
Changed Files:
_U trunk/
U trunk/src/include/k5-int.h
# Wed Jul 16 16:31:33 2008 tlyu (Taylor Yu) - Status changed from 'open' to 'review'
# Tue Jul 22 14:31:13 2008 raeburn (Ken Raeburn) - Status changed from 'review' to 'resolved'
# Fri Jan 30 23:08:19 2009 tlyu (Taylor Yu) - Version_Fixed 1.7 added