Add recipe for a centos rpm package of ScrapyD

[boogheta]

Repropose the pull request #29 without the undesired commit and with an improved postinstall test on whether scrapyd is well running.

[pablohoffman]

reStructuredText is the preferred syntax for documentation (the README and scrapy docs are all written in ReST). Can we keep that convention?

[boogheta]

done with 5f968ac

[pablohoffman]

I left a few comments.

Building a RPM for Scrapyd turns our to be a lo more work than what is proposed in this pull request, I'm afraid. To make a proper RPM, the package should declare its dependencies (twisted, scrapy, etc). Those present in CentOS (either main distro or through extras) should be used (like "python-twisted", if there's one) and the ones not present (scrapy) a RPM should be made by ourselves (that'd mean, building a RPM of Scrapy as well).

It's a different approach than what this pull request has taken, but the proper way to do it, and the only clean one that I would like to see being merged to Scrapyd repo.

I would like to hear others' opinion on the topic, @dangra what do you think?

[dangra]

Without proper dependency on an Scrapy rpm I don't see this scrapyd rpm going anywhere.

also, shouldn't the service be supervised by systemd instead of upstart on CentOS?

[boogheta]

Again, as explained before, the only dependency is scrapy (since scrapy already requires twisted, it isn't necessary to require it for scrapyd as well, it will come along the scrapy install).
If a scrapy rpm package were to be done, adding it as a dependency would indeed be the way to go.
In the mean time, the package complains when it is missing and tells you how to do it (pip install at the moment, which does not seem like possible to add as dependency in the rpm).
Right now we already spent a lot too much time on this than what we can afford, and we will use this temporary package. Maybe we can spend again some time to make a scrapy package later next year.

Regarding upstart, this is an el6 rpm, so for CentOS6, which comes with upstart, no need of SysV.

[boogheta]

I still do not really understand the reasons this was not pulled : the two concerns were on the use of upstart which as explained is fine with CentOS 6+, and on the dependencies, which are all adressed when installing scrapyd already, which can be done via pip and says so in the rpm installer. Did you guys make any progress on building a dedicated rpm for Scrapy maybe?

[jpmckinney]

Linux-flavor-specific configuration is no longer tracked in this repository. #439

Also, deployment is quite simple without packaging. #258 (comment)