Magic Blocks

Started by Simon Riggsabout 18 years ago4 messagesdocs
Jump to latest
#1Simon Riggs
simon@2ndQuadrant.com

I've just solved a problem with referencing magic blocks in plugins.

I can't see anything documented on this and I suspect others have seen
this before...
http://archives.postgresql.org/pgsql-general/2007-04/msg00908.php

While compiling with PGXS, the OBJS line said
OBJS = modulename.so

which then gave the error
ERROR: .... missing magic block

By changing the OBJS line to this
OBJS = modulename.o

then no error was received and it all works.

TFM just says "list objects files in OBJS".

It's a minor docs issue, but if y'all agree then we should add a few
more lines of wisdom in this area.

--
Simon Riggs
2ndQuadrant http://www.2ndQuadrant.com

#2Tom Lane
tgl@sss.pgh.pa.us
In reply to: Simon Riggs (#1)
Re: Magic Blocks

Simon Riggs <simon@2ndquadrant.com> writes:

While compiling with PGXS, the OBJS line said
OBJS = modulename.so

By changing the OBJS line to this
OBJS = modulename.o

then no error was received and it all works.

TFM just says "list objects files in OBJS".

Are we really in the business of teaching C programmers the difference
between object files and shared libraries?

regards, tom lane

#3Simon Riggs
simon@2ndQuadrant.com
In reply to: Tom Lane (#2)
Re: Magic Blocks

On Wed, 2008-01-23 at 19:28 -0500, Tom Lane wrote:

Simon Riggs <simon@2ndquadrant.com> writes:

While compiling with PGXS, the OBJS line said
OBJS = modulename.so

By changing the OBJS line to this
OBJS = modulename.o

then no error was received and it all works.

TFM just says "list objects files in OBJS".

Are we really in the business of teaching C programmers the difference
between object files and shared libraries?

It's a stupid mistake, but aren't most bugs, in hindsight? That's no
reason to leave it undocumented.

--
Simon Riggs
2ndQuadrant http://www.2ndQuadrant.com

#4Tom Lane
tgl@sss.pgh.pa.us
In reply to: Simon Riggs (#3)
Re: Magic Blocks

Simon Riggs <simon@2ndquadrant.com> writes:

It's a stupid mistake, but aren't most bugs, in hindsight? That's no
reason to leave it undocumented.

No, but documentation bloat is. If this were the tenth report of the
same mistake I'd be the first to agree that it was worth a paragraph.
But we have only *one* report. Furthermore, there is no reason to think
that the symptoms are consistent across platforms or PG versions, so
I don't even know what the paragraph should say.

regards, tom lane