Re: Bug: pg_regress makefile does not always copy refint.so - Mailing list pgsql-hackers

From Alvaro Herrera
Subject Re: Bug: pg_regress makefile does not always copy refint.so
Date
Msg-id 20221019062751.thsaxslbqmevki4h@alvherre.pgsql
Whole thread Raw
In response to Re: Bug: pg_regress makefile does not always copy refint.so  (Donghang Lin <donghang.lin@servicenow.com>)
List pgsql-hackers
Hi,

On 2022-Oct-18, Donghang Lin wrote:

> > I have a somewhat-related-but-not-really complaint.  I recently had need to
> > have refint.so, autoinc.so and regress.so in the install directory; but it
> > turns out that there's no provision at all to get them installed.

> The current pgsql install directory of regress only contains pg_regress binary,
> Do you suggest we add these files (excluding the scratched files) to the regress install directory?
> autoinc.so  data  expected  Makefile  parallel_schedule  pg_regress  pg_regress.c  pg_regress.h  pg_regress_main.c
README refint.so  regress.c  regressplans.sh  regress.so  resultmap  sql
 

No, I think the .c/.h files are likely included only because the RPM
rule is written somewhat carelessly.  If we add support in our
makefiles, it would have to be something better-considered.

> I think this build bug is orthogonal to the inconvenient installation/packaging.
> It produces inconsistent build result, e.g you have to run `make` twice to ensure newly built refint.so is copied to
thebuild dir.
 

Yes, I agree that it is orthogonal.  I'm not sure that what you propose
(changing these order-only dependencies into regular dependencies) is
the best possible fix, but I agree we need *some* fix.

-- 
Álvaro Herrera        Breisgau, Deutschland  —  https://www.EnterpriseDB.com/



pgsql-hackers by date:

Previous
From: vignesh C
Date:
Subject: Re: Support logical replication of DDLs
Next
From: Michael Paquier
Date:
Subject: Re: Getting rid of SQLValueFunction