Re: pgAdmin IV : Unittest modular patch

Hi Dave,

On Wed, Aug 3, 2016 at 8:32 PM, Dave Page <dave(dot)page(at)enterprisedb(dot)com>
wrote:

> On Wed, Aug 3, 2016 at 2:01 PM, Navnath Gadakh
> <navnath(dot)gadakh(at)enterprisedb(dot)com> wrote:
> > Hi Dave,
> > Thanks for clarification.
> >
> > On Wed, Aug 3, 2016 at 2:45 PM, Dave Page <dave(dot)page(at)enterprisedb(dot)com>
> > wrote:
> >>
> >> Hi Navnath
> >>
> >> On Tue, Aug 2, 2016 at 3:58 PM, Navnath Gadakh
> >> <navnath(dot)gadakh(at)enterprisedb(dot)com> wrote:
> >> > Hi Dave,
> >> > Please find the attached patch.
> >> > This patch includes:
> >> > 1. API test cases for Roles & Tablespaces node(Completed nodes: server
> >> > groups, servers, databases)
> >> > You can run test-suite using following command
> >> > for roles node
> >> > python regression/runtests.py --pkg
> >> > browser.server_groups.servers.roles
> >> > for tablespaces node
> >> > python regression/runtests.py --pkg
> >> > browser.server_groups.servers.tablespaces
> >> > for all nodes
> >> > python regression/runtests.py
> >> > You can also test with multiple servers.
> >> > 2. Delete database code in some of the missed test files.
> >> > 3. Added advanced configurations in test_advanced_config.json.in for
> >> > roles &
> >> > tablespaces. So, accordingly you need change the file
> >> > test_advanced_config.json
> >> > 4. Added one test user credentials in test_config.json.in to test the
> >> > ‘valid
> >> > password’ test case which is present in
> >> > browser/tests/test_change_password.py
> >> > Why test user credentials in test_config.json.in?
> >> > Currently, I am getting ‘UnicodeDecodeError’ when I run
> >> > test-suite(runtests.py) with existing code. I already explained the
> >> > detail
> >> > about this error in RM(#1521). I am creating test user to test the
> >> > ‘valid
> >> > password’ test case.
> >>
> >> The tablespace test is one that I think is going to cause us problems
> >> - we really can't have a hard-coded path in the config;
> >
> > Understood.
> >>
> >>
> >> - It might need to be different for each server being tested
> >> Current patch supports multiple servers if you add tablespace
> >> configurations for multiple servers. In the current patch, it's only
> for one
> >> server.
> >
> > For multiple servers you need to add multiple tablespace
> > configuration:
> > "test_tablespc_credentials":[{
> > "test_tblspace_name": "test_tablespace",
> > "test_spc_location": "/Library/PostgreSQL/9.6/data",
> > "test_spc_opts": [],
> > "test_spc_user": "XXXXX"
> > },
> > "test_tblspace_name": "test_tablespace1",
> > "test_spc_seclable": [],
> > "test_spc_location": "/Library/PostgrePlus/9.4AS/data",
> > "test_spc_opts": [],
> > "test_spc_user": "YYYYY"
> > }]
> > Anyhow, we are not going to use a hard coded path.
>
> Oh - does the per-server config override the main config? That's
> useful. So anything that's in test_advanced_config.py can be
> overridden on a per-server basis in test_config.py?

No.
per-server i.e advance config(test_advanced_config.json.in) and main
config(test_config.json) both are different files. In main config we just
mention the server’s credentials.(We can also mention per server
credentails) and in test_advanced_config.json.in(here we say per-server
config)
we mention the advanced configurations i.e. test data for each node.

>
> >> - It is very likely to be different for each user
> >>
> >> I think what we need to do is:
> >>
> >> - Skip the tests if the server is not connected via a Unix domain
> >> socket (hostname starts with /), 127.0.0.1 or ::1.
> >
> > Understood.
>
> Please output a notice saying the test was skipped, and note it in the
> summary as well (as we're planning to do with failed tests).
>
> >>
> >> - Add per-server configuration options for the tablespace path and
> >> service account under which the database server runs. These values
> >> should default to /tmp and postgres for a PostgreSQL server, and edb
> >> for PPAS.
> >
> > Ok. For server connected via Unix domain socket. As per my
> understanding
> > modified configuration should be,
> > "tablespc_credentials":[{
> > "tblspace_name": "tablespace1",
> > "testspc_seclable": [],
> > "spc_acl": [
> > {
> > "grantee":"postgres",
> > "grantor":"postgres",
> > "privileges":[
> > {
> > "privilege_type":"C",
> > "privilege":true,
> > "with_grant":false
> > }
> > ]
> > }
> > ],
> > "spc_location": "/tmp",
> > "spc_opts": [],
> > "spc_user": "postgres",
> > "service_account": "test1_user"
> > },{
> > "tblspace_name": "tablespace2",
> > "spc_seclable": [],
> > "spc_acl": [
> > {
> > "grantee":"enterprisedb",
> > "grantor":"enterprisedb",
> > "privileges":[
> > {
> > "privilege_type":"C",
> > "privilege":true,
> > "with_grant":false
> > }
> > ]
> > }
> > ],
> > "spc_location": "/tmp",
> > "spc_opts": [],
> > "spc_user": "enterprisedb",
> > "service_account": "test2_user"
> > }]
> > Please correct me if I am wrong.
>
> Something like that yes - though service_account seems pointless given
> your comment below...
>
> >> - Otherwise; assume the server is on the local machine, and:
> >> - Create /$tblspace_path/<random_string>/
> >> - chown $service_account /$tblspace_path/<random_string>/
> >> - Run the tests
> >
> > We can't set the owner for a tablespace path directory using normal
> > user, the user should be 'root'
> > Say, I am logged in my machine as a 'abc' user & I have installed
> > PostgreSQL9.5 with 'postgres' user.
> > It won't run chown postgres
> > /$tablespace_path_for_PostgreSQL9.5/<random_string>
>
> Aww crap. Yeah :-(. OK in that case let's have the user specify (and
> create) the path themselves for each server. If they don't specify a
> path, then skip the test.
>
> In fact - you might as well only skip the test in that case, and
> forget about the address check above.
>
> So, let's summarize the discussion:
- Let user specify the tablespace path in test_advanced_config.json.in
- If path not specified in the test_advanced_config.json.in skip the
test cases for tablespace.
- Output a notice saying the test was skipped, and note it in the
summary.
- No need to check the address.(Unix domain socket). It seems correct
way.

Please tell me if I missed anything.

Thanks!

--
> Dave Page
> VP, Chief Architect, Tools & Installers
> EnterpriseDB: http://www.enterprisedb.com
> The Enterprise PostgreSQL Company
>
> Blog: http://pgsnake.blogspot.com
> Twitter: @pgsnake
>

--
Thanks,
Navnath Gadakh
Software Engineer
EnterpriseDB Corporation
Mobile: +91 9975389878