Closed geissonator closed 3 years ago
I see a couple things going on here:
That line needs to use the ec overload of remote_endpoint. This is a perfect example of "Using methods that throw (or not handling bad inputs)" documented in COMMON_ERRORS.md.
Once that is fixed, your crash should go away, but you'll end up with a blank string. Next we need to figure out why the remote_endpoint call is failing. Is this test attempting to request redfish resources from within the bmc? Is it running in ipv6 mode? Is there anything else strange about this network setup (USB NIC, custom nic driver, ect) what would make remote_endpoint fail?
@sunharis Looks like this might be what you were trying to fix here: https://gerrit.openbmc-project.xyz/c/openbmc/bmcweb/+/37616
Can you comment on Andrews failure?
Thanks for the quick response @edtanous . Will need @gkeishin to comment on the test specifics but I think it basically creates a Refish Session, deletes it, then queries all sessions to be sure it's gone.
This is issue number 3 on https://github.com/openbmc/bmcweb/commit/e436008377fbcf287be02c9e9e1b59c6627d7673
Issue number 1 is theory fixed in this commit: https://gerrit.openbmc-project.xyz/c/openbmc/bmcweb/+/37580
There's another patchset here that fixes some other issues: https://gerrit.openbmc-project.xyz/c/openbmc/bmcweb/+/37616
This patch (not up yet) would fix issue number 3.
And the problem for basic auth is still there.
I think we're at the point where we need to revert e436008377fbcf287be02c9e9e1b59c6627d7673, and come at it again in a new patchset. Any objections?
We did pull in https://gerrit.openbmc-project.xyz/c/openbmc/bmcweb/+/37721 from Gunnar and our CI test passed with it. I'll let you guys sort out what you want to do though.
@edtanous : Either the upstream implementation of usage of remote_endpoint or the https://gerrit.openbmc-project.xyz/c/openbmc/bmcweb/+/37616 both are being crashed.
How about getting the IP address of the other end as of following?
accept(ssock, (struct sockaddr *)&clientAddr, &len) inet_ntoa/p(clientAddr.sin_addr).
This issue has been resolved on master for some time now. Please reopen if you can still reproduce.
We're running our CI regression suite against a downstream system. The bmcweb version used is the latest in master (f23b729676559f539790580930b1ff3b0c05805b).
During the CI test suite, we're getting this in the journal:
The test case is Delete_Redfish_Session_Using_Valid_Login. Here's the robot output leading to the fail: