From: Tom Lane <tgl@sss.pgh.pa.us>
To: Andrew Dunstan <andrew.dunstan@2ndquadrant.com>
cc: Marco Atzeri <marco.atzeri@gmail.com>, pgsql-hackers@lists.postgresql.org
Subject: Re: Cygwin linking rules
In-reply-to: <fda91d8a-93ae-d7a5-a555-4a88d5783f36@2ndQuadrant.com>
References: <1056.1538235347@sss.pgh.pa.us>
 <86419508-9d36-36a9-93ff-2a051af7d2e4@2ndQuadrant.com>
 <f82cc25e-3404-33b1-3a7b-93e5130e188e@2ndQuadrant.com>
 <257c24fc-effb-871d-4529-e3ff2ea0d436@gmail.com>
 <fda91d8a-93ae-d7a5-a555-4a88d5783f36@2ndQuadrant.com>
Comments: In-reply-to Andrew Dunstan <andrew.dunstan@2ndquadrant.com>
	message dated "Sun, 30 Sep 2018 10:13:19 -0400"
MIME-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Content-ID: <27790.1538500041.1@sss.pgh.pa.us>
Date: Tue, 02 Oct 2018 13:07:21 -0400
Message-ID: <27791.1538500041@sss.pgh.pa.us>
Precedence: bulk

Andrew Dunstan <andrew.dunstan@2ndquadrant.com> writes:
> On 09/29/2018 02:13 PM, Marco Atzeri wrote:
>> [ proposed patch ]

> Yes. So there are a couple of things here. First, the dll has 
> SO_MAJORVERSION in the name. And second it stops building any static 
> libraries and instead builds windows import libraries with names like 
> lippq.a.

I'm pretty much -1 on adding SO_MAJORVERSION to the library names.
It seems like it will cause churn to library users without really
accomplishing much, because when was the last time we changed the
SO_MAJORVERSION of anything?

I'd suggest that if we ever do change libpq's SO_MAJORVERSION,
that would be the time to append the suffix (so we'd go from libpq.dll
to libpq-6.dll).  For now, let's not fix what isn't broken.

However, the .a linking definitely is broken, and if this way
of building fixes it, that's great.  I do not have the ability
to test it, but we can throw it into the buildfarm to see what
happens.

> I think we should apply this to HEAD. If it's not too late it would 
> probably be a good thing for release 11 - would need a release note.

I think it's too late for 11; we're too close to RC1, and besides
the problem this is fixing doesn't seem to manifest before my
recent port/common library changes.  (If that's not so, maybe it
qualifies as a bug fix for back branches; but it seems rather
high risk.)

			regards, tom lane