View Revisions: Issue #4542

Summary 0004542: pubsub (gnunet-social) does not work with separated system/user configs
Revision 2016-06-09 14:21 by lynX
Description Update: This bug describes how gnunet-social is unable to relay packets to its lower layers which are running on another uid as intentional with the SYSTEM_ONLY + USER_ONLY style of configuration.

----

I wrote the "psycit" script to ensure gnunet is running in a sane
configuration and suitable egos have been generated. It prototypes
the kind of jobs that a secushare application would also have to
automate. Then I ran it on one userid to host a chatroom. It would
fetch the previously generated ego and ask gnunet-social to produce
a chatroom.

computer] ./psycit -C
=> gnunet-social -e 0A9RPQ1VDZB3M67XEBP9MVF6ZHKA4TP3YS99FMVG9JF4TQ3C9SXG --follow --host-enter --welcome
May 29 17:30:56-912320 gnunet-social-26050 WARNING Ego: 0A9RPQ1VDZB3M67XEBP9MVF6ZHKA4TP3YS99FMVG9JF4TQ3C9SXG vonlynX
May 29 17:30:56-912862 gnunet-social-26050 WARNING Ego: X33YY4YG62PQVN6918TX812ZT4EJYZAGEJDJ6STNTS2Z3HEX8WVG lynX
May 29 17:30:56-913011 gnunet-social-26050 WARNING Ego: KXM3ER7Z8J50AKFNHJ31APRKVKYPSPPA37D4SN0G7A381N4EVYE0 user
May 29 17:30:56-913174 gnunet-social-26050 WARNING Host: HXWCGDE3WBFNBE0FMFNVGMQGB3480698AHQ5Z2EEZEV6BFJ1KES0
May 29 17:30:56-913328 gnunet-social-26050 WARNING Host: GGSE8Y2RVDWM9G2W02FHJKTXC8K3MDY8Z25V81EAXNWNE9329B10

Interestingly, gnunet-social lists also previously generated
egos although we explicitly didn't want to use those, then it
outputs a new (GGSE) and an old chatroom keydress (HXWC).
After that it waits for events ("--follow").

Next I switch to a different userid on the same machine
and pass it the keydress of the newly generated place. As you
can see it quietly pulls out its own ego from gnunet-identity,
then requests a --guest-enter with TVV5 being the locally
running node.

computer ~% psycit -p GGSE8Y2RVDWM9G2W02FHJKTXC8K3MDY8Z25V81EAXNWNE9329B10 -S
=> gnunet-social --guest-enter --follow -e H85F4CKSGWF41HCFNR97E9HH6CM24P46CTS64MQQ33Z0QXMCA2K0 -p GGSE8Y2RVDWM9G2W02FHJKTXC8K3MDY8Z25V81EAXNWNE9329B10 -i TVV54YYR29KRDKM30ZC3014JXHDGPVT7KFQGA44CEV2YQSEVFDAG
May 29 17:34:29-818344 gnunet-social-26127 WARNING Ego: H85F4CKSGWF41HCFNR97E9HH6CM24P46CTS64MQQ33Z0QXMCA2K0 egal
May 29 17:34:29-818975 gnunet-social-26127 WARNING Guest: HXWCGDE3WBFNBE0FMFNVGMQGB3480698AHQ5Z2EEZEV6BFJ1KES0

The debug warnings show HXWC which probably got there by
a previous test. Next up, nothing happens. Typing anything
into stdin of any of the two processes has no effect.
Submitting a --guest-talk has no visible effects, either:

computer ~% psycit -p GGSE8Y2RVDWM9G2W02FHJKTXC8K3MDY8Z25V81EAXNWNE9329B10 -P
=> gnunet-social --guest-talk -m _notice_testing_psycit -d 'check check' -e H85F4CKSGWF41HCFNR97E9HH6CM24P46CTS64MQQ33Z0QXMCA2K0 -p GGSE8Y2RVDWM9G2W02FHJKTXC8K3MDY8Z25V81EAXNWNE9329B10 -i TVV54YYR29KRDKM30ZC3014JXHDGPVT7KFQGA44CEV2YQSEVFDAG
May 29 17:34:29-818344 gnunet-social-26127 WARNING Ego: H85F4CKSGWF41HCFNR97E9HH6CM24P46CTS64MQQ33Z0QXMCA2K0 egal
May 29 17:34:29-818975 gnunet-social-26127 WARNING Guest: HXWCGDE3WBFNBE0FMFNVGMQGB3480698AHQ5Z2EEZEV6BFJ1KES0

Hitting ^C to terminate the experiment produces the following
assertion failure in both instances of gnunet-social:

^CMay 29 17:35:35-519194 gnunet-social-26050 ERROR Assertion failed at scheduler.c:978.
child died with signal 6, without coredump

^CMay 29 17:35:39-519965 gnunet-social-26127 ERROR Assertion failed at scheduler.c:978.
child died with signal 6, without coredump
Revision 2016-05-30 07:49 by lynX
Description Excuse the summary, but it isn't quite clear if there is a bug or a misunderstanding in how this is supposed to be used.

I wrote the "psycit" script to ensure gnunet is running in a sane
configuration and suitable egos have been generated. It prototypes
the kind of jobs that a secushare application would also have to
automate. Then I ran it on one userid to host a chatroom. It would
fetch the previously generated ego and ask gnunet-social to produce
a chatroom.

computer] ./psycit -C
=> gnunet-social -e 0A9RPQ1VDZB3M67XEBP9MVF6ZHKA4TP3YS99FMVG9JF4TQ3C9SXG --follow --host-enter --welcome
May 29 17:30:56-912320 gnunet-social-26050 WARNING Ego: 0A9RPQ1VDZB3M67XEBP9MVF6ZHKA4TP3YS99FMVG9JF4TQ3C9SXG vonlynX
May 29 17:30:56-912862 gnunet-social-26050 WARNING Ego: X33YY4YG62PQVN6918TX812ZT4EJYZAGEJDJ6STNTS2Z3HEX8WVG lynX
May 29 17:30:56-913011 gnunet-social-26050 WARNING Ego: KXM3ER7Z8J50AKFNHJ31APRKVKYPSPPA37D4SN0G7A381N4EVYE0 user
May 29 17:30:56-913174 gnunet-social-26050 WARNING Host: HXWCGDE3WBFNBE0FMFNVGMQGB3480698AHQ5Z2EEZEV6BFJ1KES0
May 29 17:30:56-913328 gnunet-social-26050 WARNING Host: GGSE8Y2RVDWM9G2W02FHJKTXC8K3MDY8Z25V81EAXNWNE9329B10

Interestingly, gnunet-social lists also previously generated
egos although we explicitly didn't want to use those, then it
outputs a new (GGSE) and an old chatroom keydress (HXWC).
After that it waits for events ("--follow").

Next I switch to a different userid on the same machine
and pass it the keydress of the newly generated place. As you
can see it quietly pulls out its own ego from gnunet-identity,
then requests a --guest-enter with TVV5 being the locally
running node.

computer ~% psycit -p GGSE8Y2RVDWM9G2W02FHJKTXC8K3MDY8Z25V81EAXNWNE9329B10 -S
=> gnunet-social --guest-enter --follow -e H85F4CKSGWF41HCFNR97E9HH6CM24P46CTS64MQQ33Z0QXMCA2K0 -p GGSE8Y2RVDWM9G2W02FHJKTXC8K3MDY8Z25V81EAXNWNE9329B10 -i TVV54YYR29KRDKM30ZC3014JXHDGPVT7KFQGA44CEV2YQSEVFDAG
May 29 17:34:29-818344 gnunet-social-26127 WARNING Ego: H85F4CKSGWF41HCFNR97E9HH6CM24P46CTS64MQQ33Z0QXMCA2K0 egal
May 29 17:34:29-818975 gnunet-social-26127 WARNING Guest: HXWCGDE3WBFNBE0FMFNVGMQGB3480698AHQ5Z2EEZEV6BFJ1KES0

The debug warnings show HXWC which probably got there by
a previous test. Next up, nothing happens. Typing anything
into stdin of any of the two processes has no effect.
Submitting a --guest-talk has no visible effects, either:

computer ~% psycit -p GGSE8Y2RVDWM9G2W02FHJKTXC8K3MDY8Z25V81EAXNWNE9329B10 -P
=> gnunet-social --guest-talk -m _notice_testing_psycit -d 'check check' -e H85F4CKSGWF41HCFNR97E9HH6CM24P46CTS64MQQ33Z0QXMCA2K0 -p GGSE8Y2RVDWM9G2W02FHJKTXC8K3MDY8Z25V81EAXNWNE9329B10 -i TVV54YYR29KRDKM30ZC3014JXHDGPVT7KFQGA44CEV2YQSEVFDAG
May 29 17:34:29-818344 gnunet-social-26127 WARNING Ego: H85F4CKSGWF41HCFNR97E9HH6CM24P46CTS64MQQ33Z0QXMCA2K0 egal
May 29 17:34:29-818975 gnunet-social-26127 WARNING Guest: HXWCGDE3WBFNBE0FMFNVGMQGB3480698AHQ5Z2EEZEV6BFJ1KES0

Hitting ^C to terminate the experiment produces the following
assertion failure in both instances of gnunet-social:

^CMay 29 17:35:35-519194 gnunet-social-26050 ERROR Assertion failed at scheduler.c:978.
child died with signal 6, without coredump

^CMay 29 17:35:39-519965 gnunet-social-26127 ERROR Assertion failed at scheduler.c:978.
child died with signal 6, without coredump