Ludovic Courtès <ludo@gnu.org> writes:

> Hello,
>
> swedebugia <swedebugia@riseup.net> skribis:
>
>> On 2018-11-04 11:44, Christopher Baines wrote:
>>> ---
>>>   gnu/services/web.scm | 297 ++++++++++++++++++++++++++++++++++++++++++-
>>>   gnu/tests/web.scm    | 104 ++++++++++++++-
>>>   2 files changed, 399 insertions(+), 2 deletions(-)
>>
>> Nice work with the service-declaration. :) (untested)
>
> +1!
>
>> Would you be willing to update the patch with documentation also?
>
> Yes, also with ‘documentation’ fields.  :-)
>
> I spotted a typo here:
>
> +(define-record-type* <patchwork-configuration>
> +  patchwork-configuration make-patchwork-configuration
> +  patckwork-configuration?
>       ^^

Good spot!

> Regarding ‘patchwork-setup-gexp’, I wonder if you could use
> ‘make-forkexec-constructor’ with the appropriate environment variables
> and move the “createuser” bit to an activation snippet.
>
> Hmm maybe the activation snippet would run too early, right?  In that
> case, perhaps you could create another Shepherd service,
> ‘patchwork-initialization’, that would do the createuser stuff, and have
> ‘patchwork’ depend on it.

So, I've made some changes since I last sent this patch, the biggest
being splitting the database creation out from running of the database
migrations.

Assuming that the shepherd service defined as part of the
patchwork-service-type just runs the migrations, yes, it needs
PostgreSQL (or whatever database you're using) to be available.

At the moment, the service is hardcoded to use mod_wsgi, so it runs
through the httpd-service, no shepherd service for running patchwork is
needed in the service-type at least.

It would be good to try and make it more flexible in the future, at
least so that you can pick and choose a bit more (and instead use uwsgi,
gunicorn, ...).

Thanks for taking a look :)

Chris