Thread: [HACKERS] Allowing nonzero return codes from \quit
Another suggestion to come out of the \quit_if discussion was that someone might want to send a return code other than 0 back. Something like this:
\quit 4or even:
This isn't a personal need of mine, but I figured it was an idea worth discussing on its own.
\set exit_code 127
\quit :exit_code
2017-01-23 19:09 GMT+01:00 Corey Huinker <corey.huinker@gmail.com>:
Another suggestion to come out of the \quit_if discussion was that someone might want to send a return code other than 0 back. Something like this:\quit 4or even:This isn't a personal need of mine, but I figured it was an idea worth discussing on its own.
\set exit_code 127
\quit :exit_code
\quit exit_code is better - if we define some special variable, then we have to specify when it should be used and when not. Taking value from command is clean without any another questions.
Regards
Pavel
>> \quit 4 As \q does not currently have an argument, this seems an easy and reasonnable extension. However, currently there are 4 existing exit status for psql: 0 (ok), 1 (fatal error), 2 (connection error), 3 (script error...). +128 status are also already used when killing a psql process. ISTM that quitting with a status should not interfere with these cases at least, because a shell could not rely on the exit status to know what went wrong? Now having \q 1/2/3 forbidden would also be a strange behavior... >> \set exit_code 127 >> \quit :exit_code >> >> This isn't a personal need of mine, but I figured it was an idea worth >> discussing on its own. > > \quit exit_code is better - if we define some special variable, then we > have to specify when it should be used and when not. Taking value from > command is clean without any another questions. With minimal luck the second form would probably work out of box if "\quit <int>" is implemented because of the way variable substitutions are performed. -- Fabien.
On Mon, Jan 23, 2017 at 1:26 PM, Fabien COELHO <coelho@cri.ensmp.fr> wrote:
\quit 4
As \q does not currently have an argument, this seems an easy and reasonnable extension.
However, currently there are 4 existing exit status for psql: 0 (ok), 1 (fatal error), 2 (connection error), 3 (script error...). +128 status are also already used when killing a psql process.
I didn't think about it too much, but I don't see why a user couldn't set one of those error codes.
I did, however, think that any attempt to set an exit_code outside of [0,127] would itself be an error, resulting in an exit code of 3.
I did, however, think that any attempt to set an exit_code outside of [0,127] would itself be an error, resulting in an exit code of 3.
ISTM that quitting with a status should not interfere with these cases at least, because a shell could not rely on the exit status to know what went wrong? Now having \q 1/2/3 forbidden would also be a strange behavior...\set exit_code 127
\quit :exit_code
This isn't a personal need of mine, but I figured it was an idea worth
discussing on its own.
\quit exit_code is better - if we define some special variable, then we
have to specify when it should be used and when not. Taking value from
command is clean without any another questions.
With minimal luck the second form would probably work out of box if "\quit <int>" is implemented because of the way variable substitutions are performed.
Yeah, sorry if I wasn't clear on that. I wasn't proposing a special variable named exit_code. I was just showing the setting of a regular psql variable.
>> As \q does not currently have an argument, this seems an easy and >> reasonnable extension. >> >> However, currently there are 4 existing exit status for psql: 0 (ok), 1 >> (fatal error), 2 (connection error), 3 (script error...). +128 status are >> also already used when killing a psql process. > > I didn't think about it too much, but I don't see why a user couldn't set > one of those error codes. > I did, however, think that any attempt to set an exit_code outside of > [0,127] would itself be an error, resulting in an exit code of 3. Hmmm. Maybe it should let the user shoots its own foot if desired, just a reminder in the doc of a possible interference with existing codes would be enough? -- Fabien.
Fabien COELHO <coelho@cri.ensmp.fr> writes:
>> I didn't think about it too much, but I don't see why a user couldn't set
>> one of those error codes.
>> I did, however, think that any attempt to set an exit_code outside of
>> [0,127] would itself be an error, resulting in an exit code of 3.
> Hmmm. Maybe it should let the user shoots its own foot if desired, just a
> reminder in the doc of a possible interference with existing codes would
> be enough?
Yeah, I see no reason why scripts shouldn't be allowed to return one of
those codes if they like. We do need to constrain it to [0,127] for
sanity though, or possibly [0,125] --- doesn't POSIX mandate special
meaning for 126 and 127?
regards, tom lane