Thanks, yes a stacktrace from gnome-keyring-daemon is necessary. 

FWIW, UI programs should generally use asynchronous gnome-keyring functions rather than the synchronous ones (which evolution is doing in the stack trace above). This allows them to keep the UI interactive during a gnome-keyring-daemon prompt or some such.

Point taken, although Evolution's password management is already asynchronous, not to mention complex and brittle.  I hesitate to pile another asynchronous layer on top.  There's a case to be made that our whole password management design needs to be rethought, but my previous attempts have failed.  That code just hates me.

Anyway, wanted to mention that a lot of Evolution users also report seeing this filling up their /var/log/messages during the hang:

   Jun 15 09:11:55 localhost gnome-keyring-daemon[2607]: couldn't read 4 bytes
   from client:

Also, the hang has been reported on both 32 and 64-bit architectures.

and another comment -- these last few days I have not experienced any hangs. Before that, I would get at least one hang every day.

BUT -- I *did* change my Evo account setup, raising the interval between POP mail checks from 5 to 10 minutes (I was wondering if this could be playing a role...). I will reduce it back to 5 minutes (or even lower), and try again.

There was an erroneous warning (ie: 'couldn't read 4 bytes from client') has been fixed in recent versions of gnome-keyring. See bug #511285

OK, got it again.

Evo BT:

+ Trace 204608

and g-kr-daemon's bt:

(gdb) thread apply all bt full

+ Trace 204609

Thanks for the hard work in getting the stack trace. This is going to be a tough one to figure out. But I'll give it a shot. How often can you duplicate this problem?

Stef, usually once per day, when I had email checking every 5 minutes or so.  Right now, with a longer interval, it is once per week or so.

I will downset the interval again, and we will see what happens. If you would like me to to something specific, please tell me.

*** Bug 548845 has been marked as a duplicate of this bug. ***

Created attachment 117170 [details] [review]
Patch to try

Does this patch fix the problem you're experiencing?

The patch does not apply successfully on 2.23.90, but applies successfuly on trunk (revision 1260); but it fails to build, on gp11, so it has nothing to do with the patch:

libtool: compile:  gcc -DHAVE_CONFIG_H -I. -I.. -I.. -I.. -I/usr/include/glib-2.0 -I/usr/lib/glib-2.0/include -pthread -I/usr/include/glib-2.0 -I/usr/lib/glib-2.0/include -I/usr/include/glib-2.0 -I/usr/lib/glib-2.0/include -Wall -Wchar-subscripts -Wmissing-declarations -Wmissing-prototypes -Wnested-externs -Wpointer-arith -Wcast-align -Wsign-compare -Werror -g -O2 -Wno-strict-aliasing -Wno-sign-compare -MT gp11-call.lo -MD -MP -MF .deps/gp11-call.Tpo -c gp11-call.c  -fPIC -DPIC -o .libs/gp11-call.o
cc1: warnings being treated as errors
gp11-call.c: In function ‘_gp11_call_sync’:
gp11-call.c:355: error: format not a string literal and no format arguments
gp11-call.c: In function ‘_gp11_call_basic_finish’:
gp11-call.c:444: error: format not a string literal and no format arguments
make[4]: *** [gp11-call.lo] Error 1
make[4]: Leaving directory `/usr/src/buildd/gnome-keyring-svn/gp11'
make[3]: *** [all-recursive] Error 1
make[3]: Leaving directory `/usr/src/buildd/gnome-keyring-svn/gp11'
make[2]: *** [all] Error 2
make[2]: Leaving directory `/usr/src/buildd/gnome-keyring-svn/gp11'
make[1]: *** [all-recursive] Error 1
make[1]: Leaving directory `/usr/src/buildd/gnome-keyring-svn'
make: *** [all] Error 2
 
Base system is Ubuntu Alpha4, Gnome 2.23.90.

I am unsure if there are special instructions to build g-kr, and I could not find a site talking about it. Or, perhaps, I should not apply the patch to trunk?

Thanks for letting me know. I've fixed that build problem on trunk. Although I didn't get those build errors, the problems they were pointing to were legitimate.

Stef, some more:

cc1: warnings being treated as errors
gkr-crypto.c: In function ‘fatal_handler’:
gkr-crypto.c:63: error: format not a string literal and no format arguments
gkr-crypto.c: In function ‘gkr_crypto_sexp_dump’:
gkr-crypto.c:808: error: format not a string literal and no format arguments
make[3]: *** [libgkr_common_la-gkr-crypto.lo] Error 1
make[3]: Leaving directory `/usr/src/buildd/gnome-keyring-svn/common'
make[2]: *** [all-recursive] Error 1

Cool, your compiler found some possible bugs. What compiler are you using? 

Compile errors fixed: 

2008-08-27  Stef Walter  <stef@memberwebs.com>

	* common/gkr-crypto.c: Fix build problems with string formats.

gcc 4.3.1, Ubuntu Intrepid. Will try it again :-)

Yet another:

cc1: warnings being treated as errors
gkr-wakeup.c: In function ‘gkr_wakeup_now’:
gkr-wakeup.c:93: error: ignoring return value of ‘write’, declared with attribute warn_unused_result
make[3]: *** [libgkr_common_la-gkr-wakeup.lo] Error 1
make[3]: Leaving directory `/usr/src/buildd/gnome-keyring-svn/common'

notice the first line. This is a make from SVN, so the compiler options are as SVN has them set. I simply took out -Werror from the Makefile under ./common, and I was then able to build g-kr. Testing now.

That's a silly warning from gcc. I've committed a hopeful fix to silence the warning.

As far as -Werror .... We use -Werror in SVN checkout versions to catch problems that we wouldn't otherwise find, as the gnome-keyring maintainers obviously don't have every compiler, OS, and distro at their disposal.

I have been running your patch for a while, and have not had another lock.

I would guess this is a go... of course, this is a deadlock, and it might just be I did not hit the right spot so far. But, given that I would be hit by it almost every day, not having had it for about 2 weeks is a good indication.

Thank you.

*** Bug 558934 has been marked as a duplicate of this bug. ***

Great let's close this.