Re: [Yadab-devel] Project/Client to html page

[Henrik Sandklef]

On 04/09/2011 11:32 PM, Patrik Willard wrote:
> On 04/09/2011 10:47 PM, Henrik Sandklef wrote:
>> On 04/08/2011 12:36 AM, Patrik Willard wrote:

>> In the example above, the project1 and client1 are directories and
>> needed to create links to the actual (text) reports in the table on the
>> daily report page. Like this:
> 
> When I think about it, what is stopping us from "normalizing" the
> project and client names (replacing everything not [a-zA-Z0-9-_] with
> "_" or whatever) and do away with the "projectX" "clientY" (where X and
> Y are positive integers) altogether?

I would like the:

        project names
        client names

to be human readable and mean something. For example a project called:

        Xnee from CVS HEAD

actually gives hint about what is built. And a client name such as:

        Debian Lenny ppc

gives at least some hints on what type of computer it is.


The names are thus META information where the "project1" would be
the.... well, container/directory of the project info.

> I am pretty sure that I have already built the database to cope with
> such an event (with the exception that if two people give their systems
> an identical hostname, they will be identified as the same client, which
> probably is unsatisfactory.)

DAMN DAMN DAMN DAMN ... I have forgotten about multiple build hosts.

Below is a draft sketch (tree -d)

buildlogs/
`-- 2011-04-01
    `-- hosts
        |-- host1
        |   `-- projects
        |       |-- project1
        |       |   `-- clients
        |       |       |-- client1
        |       |       |-- client2
        |       |       |-- client3
        |       |       `-- client4
        |       `-- project2
        |           `-- clients
        |               |-- client1
        |               |-- client2
        |               |-- client3
        |               `-- client4
        `-- host2
            `-- projects
                |-- project1
                |   `-- clients
                |       |-- client1
                |       |-- client2
                |       |-- client3
                |       `-- client4
                `-- project2
                    `-- clients
                        |-- client1
                        |-- client2
                        |-- client3
                        `-- client4




> (How does yadab handle that right now? Say if I set up a build-client,
> how would the host go about assigning me a client number?)
> 
>> <td class="setup ok"><a
>> href="buildlogs/2011-04-01/projects/project1/clients/client2/setup.txt">OK</a></td>
>> <td class="unpack ok"><a
>> href="buildlogs/2011-04-01/projects/project1/clients/client2/unpack.txt">OK</a></td>
>> <td class="prepare ok"><a
>> href="buildlogs/2011-04-01/projects/project1/clients/client2/prepare.txt">OK</a></td
> 
> Off-topic: oh how optimistic of you ("<td class="... *ok*">) ;)
> 
>> We (imho) also need to add info on OS, ARCH. This can be done using the
>> sys-info-long.txt file. So we might as well get the client name (NODE)
>> info from the same file, so I suggest to remove
> 
>>              # Client name
> 
>> from the projects-results.txt file. New "look":
> 
> 
>>>> # Project directory
>>>> # Client directory
>>>> # Project name
>>>> # Connection status
>>>> # Return value from setup
>>>> # Return value from unpack
>>>> # Return value from prepare
>>>> # Return value from clean
>>>> # Return value from build
>>>> # Return value from check
>>>> # Return value from install
> 
> 
> I support this :)
> 
> Just to verify that I am on the same wavelength: the project name should
> also go into the database (alongside the project directory), as well as
> architecture, hostname and OS (with the client), right?

I think a note here is important:

All the build logs (txt files) should be up there on the web server as
well as the yadab reports. The log files (txt) contain the real/raw
information about the builds where as yadab presents them in a more
human friendly way.

Given that, we can always find the information afterwards.

But I think that it would be cool to (later on be able to search for
things like:
        all builds on x86_64 machines
so yes, put them in db (imho).


/crapface