Re: postgresql HEAD build failure

Alvaro Herrera wrote:
> John Summerfield wrote:
>> I have, for some months, been running this daily, but not taking notice
>> of the results:
>> [pgtest(at)numbat ~]$ cat bin/pg_build
>> #!/bin/bash
>> set -xe
>> cd /home/pgtest/pgsql/pgsql-2007-04-23/
>> ping -c4 2>/dev/null anoncvs.postgresql.org | grep -q 'bytes from'\
>> && {
>> make distclean || :
>> cvs -q update
>> ./configure --prefix=${HOME}//postgres --with-pgport=5433 --with
>> -pam --with-openssl \
>> --enable-depend \
>> --with-gnu-ld --with-python --with-perl
>> }
>> time make -j3 -s && time make -j3 -s install
>> [pgtest(at)numbat ~]$
>>
>> It fails thus:
>> config.status: linking ./src/makefiles/Makefile.linux to src/Makefile.port
>> + make -j3 -s
>> tsearchcmds.c:38:30: error: tsearch/ts_cache.h: No such file or directory
>> tsearchcmds.c:39:31: error: tsearch/ts_public.h: No such file or directory
>> tsearchcmds.c:40:30: error: tsearch/ts_utils.h: No such file or directory
>
> Hmm, try deleting the build tree and start again. Sometimes, what
> happens to me is that somebody adds a new directory, and my build tree
> does not contain it. It's easy to fix: when I had a slower computer
> what I did was create the offending directory and symlink the makefile.
> Nowadays I just wipe the whole thing and let the build directory be
> constructed again by configure. That gets things in sync.

that, I think, subverts the purpose of cvs and other SCMTs, but...
>
> Now that I look closer, however, this was always with source dirs, not
> include dirs. I think the problem here is that you're not passing -d to
> "cvs update".
>
-d fixed the missing directories
and -P prunes obsolete ones.

Obviously I needed to read the instructions more closely:-(

And the timing of my upgrade was just a coincidence. It happens.

Thanks for your help.