On 06/ 3/10 05:30 PM, [email protected] wrote:
On Thu, Jun 03, 2010 at 05:27:25PM -0700, Shawn Walker wrote:
On 06/ 3/10 05:21 PM, [email protected] wrote:
On Thu, Jun 03, 2010 at 05:18:58PM -0700, Shawn Walker wrote:
On 06/ 3/10 05:12 PM, [email protected] wrote:
Folks,
This fixes a bug in publication that's caused by a problem in the NFS4
client implementation. (See: 4877567). The solution for pkg is to open
all manifests as O_RDWR in the publication path, instead of some
O_RDONLY and some O_WRONLY.
http://cr.opensolaris.org/~johansen/webrev-13194/
This seems right to me. Since an automated test case isn't
practical, I'll assume that this was verified to have resolved the
issue manually.
I wrote a test program that reproduces the problem, and when changed in
the same way that this is, the EINVAL disappeared. If someone who's
hitting this problem wants to patch their build and see if this solves
the problem, that would help us verify the fix prior to putback.
I was actually referring to having a unit test in our gate, but it's
nice to know an automated one was possible outside of it :)
Er, sorry. I can test that the NFS bug is present, and that changing
the way we open files solves the problem. I have yet to figure out how
to write a concise test that actually uses the code in Transaction.py,
if that's what you're asking?
Yes, a test of the fix as applied to publication. I suppose the only
way to do that is to find someone having this problem consistently.
Cheers,
-Shawn
_______________________________________________
pkg-discuss mailing list
[email protected]
http://mail.opensolaris.org/mailman/listinfo/pkg-discuss
