Re: psql \ir filename normalization

From: Robert Haas <robertmhaas(at)gmail(dot)com>
To: Josh Kupershmidt <schmiddy(at)gmail(dot)com>
Cc: pgsql-hackers <pgsql-hackers(at)postgresql(dot)org>, Gurjeet Singh <singh(dot)gurjeet(at)gmail(dot)com>
Subject: Re: psql \ir filename normalization
Date: 2011-11-16 04:28:37
Message-ID: CA+TgmoZnszt86+GdFNk8VX5L8xDt4M+K+vUzdd=CV+_g_N14Zg@mail.gmail.com
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-hackers

On Tue, Nov 15, 2011 at 6:54 PM, Josh Kupershmidt <schmiddy(at)gmail(dot)com> wrote:
> Commit c7f23494c1103f87bcf1ef7cbfcd626e73edb337 editorialized a bit on
> Gurjeet Singh's patch to implement \ir for psql, particularly in
> process_file(). Unfortunately, it looks like it broke the common case
> of loading a .SQL file in psql's working directory. Consider the
> following test case:
>
> mkdir -p /tmp/psql_test/subdir/
> mkdir -p /tmp/psql_test/path2/
>
> echo "SELECT 'hello 1';" > /tmp/psql_test/hello.sql
> echo "SELECT 'hello from parent';" > /tmp/psql_test/hello_parent.sql
> echo "SELECT 'hello from absolute path';" >
> /tmp/psql_test/path2/absolute_path.sql
> echo -e "SELECT 'hello 2';\n\ir ../hello_parent.sql\n\ir
> /tmp/psql_test/path2/absolute_path.sql" >
> /tmp/psql_test/subdir/hello2.sql
> echo -e "\ir hello.sql\n\ir subdir/hello2.sql" > /tmp/psql_test/load.sql
>
>
> If you try to load in "load.sql" from any working directory other than
> /tmp/psql_test/ , you should correctly see four output statements.
> However, if you:
>  cd /tmp/psql_test/ && psql test -f load.sql
>
> You will get:
>
> psql:load.sql:1: /hello.sql: No such file or directory
> psql:load.sql:2: /subdir/hello2.sql: No such file or directory
>
> Attached is a patch which fixes this, by recycling the bit of
> Gurjeet's code which used "last_slash". (I have a feeling there's a
> simpler way to fix it which avoids the last_slash complications.)

Argh. The root of the problem here seems to be that
join_path_components() feels entitled to arbitrarily insert a pathname
separator at the front of the output string even if its first input
didn't begin with one originally. Lame!

While looking for other places where this behavior might cause a
problem, I noticed something else that doesn't seem right. On
REL9_1_STABLE, if I initdb and then change the first "all" on the
"local" line to "@foo", I get this:

LOG: could not open secondary authentication file "@foo" as
"/Users/rhaas/pgsql/x/foo": No such file or directory

...and then the server starts up. But on master, I get:

LOG: could not open secondary authentication file "@foo" as
"/Users/rhaas/pgsql/y/foo": No such file or directory
LOG: end-of-line before database specification
CONTEXT: line 84 of configuration file "/Users/rhaas/pgsql/y/pg_hba.conf"
LOG: invalid connection type "all"
CONTEXT: line 85 of configuration file "/Users/rhaas/pgsql/y/pg_hba.conf"
FATAL: could not load pg_hba.conf

...which doesn't look right.

--
Robert Haas
EnterpriseDB: http://www.enterprisedb.com
The Enterprise PostgreSQL Company

In response to

Responses

Browse pgsql-hackers by date

  From Date Subject
Next Message Greg Smith 2011-11-16 04:53:53 Configuration include directory
Previous Message Tom Lane 2011-11-16 03:40:49 Re: ISN was: Core Extensions relocation