Re: Extensions, this time with a patch - Mailing list pgsql-hackers

From Itagaki Takahiro
Subject Re: Extensions, this time with a patch
Date
Msg-id AANLkTin7=xegS=uYFTr6SPFp63ctyR9o_VzUgsUFNR5o@mail.gmail.com
Whole thread Raw
In response to Re: Extensions, this time with a patch  (Dimitri Fontaine <dimitri@2ndQuadrant.fr>)
Responses Re: Extensions, this time with a patch
List pgsql-hackers
On Thu, Oct 21, 2010 at 7:12 AM, Dimitri Fontaine
<dimitri@2ndquadrant.fr> wrote:
> This control file contains at minimum a single line for the name of the
> extension, but it's better already with a comment for users. I've been
> filling them for our extensions, pasting from the documentation:
>
>         name        | version  |
> --------------------+----------+
>  fuzzystrmatch      | 9.1devel |
>  hstore             | 9.1      |

Why does only hstore have version '9.1'? Any other modules have '9.1devel'.

> If you provide a $(VERSION) variable, then a line in the control file is
> automatically added at make install (version = '$(VERSION)'), in order
> to make life easier for extension authors.

In v7, a line of "version = '...'" is added at "make install", and removed
at "make clean". Also, if we runs "make install" multiple times, version
lines are added repeatedly. I don't think they are good ideas; we should
not modify source codes stored in git repo when we build them.

How about having *.control.in and replace magic keywords in them at "make"?
"make install" won't modify files at all, and "make clean" just removes
*.control. It is the way we're using for *.sql.in and MODULE_PATHNAME.

> Some extensions are missing here because they fail to compile on my
> workstation where all the libs aren't installed --- ossp, xml2, etc

I found xml2/pgxml.control should have 'pgxml" for the name.

--
Itagaki Takahiro


pgsql-hackers by date:

Previous
From: Robert Haas
Date:
Subject: Re: max_wal_senders must die
Next
From: Robert Haas
Date:
Subject: Re: default_statistics_target WAS: max_wal_senders must die