Minor bug: Odd feedback on STDERR from PSQL for block comments
Guys,
Here's a non-showstopper in 7.4RC1, but probably good to fix before RC2:
(btw, tested & exists on beta4 as well, so I'm not sure when it was
introduced)
1) Create a file with *only* a block comment in it, or with a block comment as
the very last line.
2) run psql -f filename database
You will get some odd feedback on STDERR:
on Linux, file with only a block comment:
josh@temoku:~/Documents/oss/postgres> psql -f tempit2.sql -U postgres test1
psql:tempit2.sql:1: josh@temoku:~/Documents/oss/postgres>
on OSX:
josh% /usr/local/pgsql/bin/psql -U postgres -f ~/test.sql sharky
psql:/Users/josh/test.sql:1: [mercury:~] josh%
on Linux, file with several commands, ending in a block comment:
josh@temoku:~/Documents/oss/postgres/phpcon> psql -f tempit.sql -U postgres
test1
CREATE TABLE
INSERT 17168 1
INSERT 17169 1
INSERT 17170 1
psql:tempit.sql:13: ERROR: column "iddomain" does not exist
DELETE 1
psql:tempit.sql:18: josh@temoku:~/Documents/oss/postgres/phpcon>
I'm not sure why we're getting a line header ("psql:tempit2.sql:1: ") on an
empty line, but it's annoying for any application writer (Bricolage) looking
for failure messages on STDERR.
--
-Josh Berkus
Aglio Database Solutions
San Francisco
Josh Berkus <josh@agliodbs.com> writes:
Here's a non-showstopper in 7.4RC1, but probably good to fix before RC2:
1) Create a file with *only* a block comment in it, or with a block comment as
the very last line.
2) run psql -f filename database
You will get some odd feedback on STDERR:
I can confirm the problem. It happens in more cases than Josh mentions,
as well:
-- running the file with \i also shows the problem
-- there can be blank lines and whitespace after the
block comment, and it still shows up.
Example: z.sql contains
--- snip ---
select 2+2;
/* block */ -- z
--- snip ---regression=# \i z.sql
?column?
----------
4
(1 row)
psql:z.sql:5: regression=#
Peter, do you have time to look at this? You're probably more familiar
than I am with the psql command loop.
regards, tom lane
Guys,
I can confirm the problem. It happens in more cases than Josh mentions,
as well:
-- running the file with \i also shows the problem
-- there can be blank lines and whitespace after the
block comment, and it still shows up.
This bug is still present in RC2. Are we going to fix it before release?
According to David W., it really screws up the error-detection code on
Bricolage, and I'd think that other applications might have the same problem.
--
-Josh Berkus
Aglio Database Solutions
San Francisco
Josh Berkus <josh@agliodbs.com> writes:
This bug is still present in RC2. Are we going to fix it before release?
I have a fix (stupid case omission in AcceptResult()) but I am having
some trouble reaching the CVS server right at the moment ...
regards, tom lane