URL:
https://github.com/SSSD/sssd/pull/128
Title: #128: Fix group renaming issue when "id_provider = ldap" is set
fidencio commented:
"""
Okay, I've been playing with the suggestion but I'm kind stuck with one problem
here: how to invalidate the negative cache from the sysdb_ops.c?
@pbrezina has suggested to just fire a d-bus call to nss responder and do it from there,
which is a quite good option (and I have it partially implemented) but in order to
communicate with the client, well, I need access to the struct dp_client ... which
doesn't seem something easy to have from the sysdb_ops.
@sumit-bose suggested to add a new return error (something like
SUCCESS_BUT_RENAME_DETECED) and treat it way up in the stack and then update the memory
cache.
Both those ideas are quite intrusive and the second, somehow, seems more error-prone than
the other.
There's also the possibility to start passing down the id_ctx (which has a pointer to
be_ctx, which has a pointer to data_provider, from where I can get the struct dp_client)
... so we can just fire the d-bus call in the right place ... but that's also
something that is quite ugly to do.
So, I'm really open for suggestions here ...
"""
See the full comment at
https://github.com/SSSD/sssd/pull/128#issuecomment-366279844