On Fri, Oct 28, 2011 at 1:24 PM, Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> wrote:
> Robert Haas <robertmhaas(at)gmail(dot)com> writes:
>> On my Fedora 14 box, make -j3 fails. I think the below is the
>> relevant portion of the output. This has a passing similarity to bug
>> 5665, but I can't for the life of me see what the problem is here.
>> parer.o depends on preproc.h, which depends on preproc.c; therefore,
>> parser.o should not be built until preproc.c has been built, but
>> that's exactly what make is doing.
> I tried to reproduce this on my own Fedora 14 box, and couldn't.
> I cd'd to src/interfaces/ecpg/preproc and did several repetitions of
> make maintainer-clean
> make -j
> and every time, make carefully waited until bison was done before
> launching the compiles of preproc.o, parser.o, and the other files
> that are declared to depend on preproc.h.
I see the same thing. But when I do make -j3 at the toplevel, it
bombs out as described before.
> I *can* reproduce failures if I do the same thing one directory level
> up, in src/interfaces/ecpg. But those are caused by the other sub-makes
> not waiting for include/ecpg_config.h to get built. Fixing that is
> beyond my level of make-fu.
I can also reproduce this problem. I think this one does not occur on
a fresh build because ecpg_config.h is created by configure and is
only removed by make maintainer-clean. If I make maintainer-clean in
src/interfaces/ecpg, then rerun config.status, and then do make -j it
succeeds. However, if I repeat those steps from one level further up,
in src/interfaces, then it bombs out as described in my OP.
> This is with make-3.82-3.fc14.x86_64 ...
That's the same make I'm using here.
The Enterprise PostgreSQL Company
In response to
pgsql-hackers by date
|Next:||From: Tom Lane||Date: 2011-10-28 18:37:43|
|Subject: Re: TOAST versus VACUUM, or "missing chunk number 0 for toast value" identified |
|Previous:||From: Robert Haas||Date: 2011-10-28 17:58:51|
|Subject: Re: So, is COUNT(*) fast now?|