On 2020-07-14 14:38, Steve Dickson wrote:
gssd: Refcount struct clnt_info to protect multithread usage
Update to libevent 2.x apis.
gssd: Cleanup on exit to support valgrind.
gssd: gssd_k5_err_msg() returns a strdup'd msg. Use free() to release.
gssd: Fix locking for machine principal list
gssd: Add a few debug statements to help track client_info lifetimes.
gssd: Lookup local hostname when srchost is '*'
gssd: We never use the nocache param of gssd_check_if_cc_exists()
Fix various clang warnings.
I did commit all of the above... (tag: nfs-utils-2-5-2-rc1)
Oops, I'd been working on an updated patch set. There's nothing really
actively wrong in the above patches. I'd just gone back and added proper
NULL checks and error messages for the libevent conversion. I'll rebase
and send that as an update. Also "gssd: Lookup local hostname when
srchost is '*'" I think is wrong. After the first day I couldn't get it
to repeat, and think it was a mis-compile issue. However, the new code
arrangement is better in that it shows the correct dns name
transformation. Previously it would use the same buffer for input &
output which made the log message very confusing. I'll just drop the '*'
check.
I did not commit the following
Cleanup printf format attribute handling and fix format strings
because 3 different version were posted
Cleanup printf format attribute handling and fix various format strings
Cleanup printf format attribute handling and fix format strings
Consolidate printf format attribute handling and fix various format strings
I was not sure which one you wanted and I was wondering what exactly is
being cleaned up? What problems is this solving?
They're all the same patch. The summary line was wrapping in the cover
letter so I edited it a few times, not realizing that format-patch was
creating another file even if I aborted.
So, it actually does a few things, all based around fixing printf style
formats.
There were 2 different macros defined to add printf format attribute to
functions, and several open codings. So it first consolidates them into
one set of macros (although there is a second copy in nfsidmap.h since
that's an installed file and can't depend on config.h.
Then, there were several functions that were not marked with the printf
format attribute (nfsidmap plugins and gssd printerr()).
Finally, a cleanup of all the resulting gcc & clang warnings on both 32
& 64 bit. In several cases some real errors, not enough parameters,
passing in various types for the dynamic length which requires an int,
passing in a char** instead of char*, etc. Of course these are mainly
debugging messages so rarely caused an issue but were in need of
cleaning up.
Finally, being this is a whole tree commit and I have a number
of patches in the queue.. I would like to hold off on this one.
A patch like this will cause all those patches in the queue
not to apply... So once I drain the queue, hopefully you
would not mind rebasing... after we talk about what you
are trying to do.
Not a problem. I have it rebased here and can send it at any time, or
split it up if you prefer.
I do appreciate the hard work... esp with gssd... I did test
it every step of the way... and it seems to be fairly
solid... nice work!
I've been chasing that threading bug for over a year. Trying to stress
the number of simultaneous mounts, types, etc. never thinking the issue
was external. I bet if I went back and correlated the crashes I saw,
probably happened when I was upgrading or rebooting the kdc.
Doug
