psql 'none' as a HISTFILE special case

On Mon, 2006-08-21 at 19:27 +0300, Martin Atukunda wrote:

If readline is used by psql, a history file is automatically used.
This patch adds the special file name 'none', which if set as the
HISTFILE parameter, will cause psql not to use the history file.

I think it would be cleaner to use a separate \set variable to control
whether a history file is written, rather than needlessly overloading
the meaning of HISTFILE.

-Neil

Neil Conway <neilc@samurai.com> writes:

On Mon, 2006-08-21 at 19:27 +0300, Martin Atukunda wrote:

If readline is used by psql, a history file is automatically used.
This patch adds the special file name 'none', which if set as the
HISTFILE parameter, will cause psql not to use the history file.

I think it would be cleaner to use a separate \set variable to control
whether a history file is written, rather than needlessly overloading
the meaning of HISTFILE.

Why is this useful at all? There's already the -n (don't use readline)
switch.

regards, tom lane

Tom Lane wrote:

Neil Conway <neilc@samurai.com> writes:

On Mon, 2006-08-21 at 19:27 +0300, Martin Atukunda wrote:

If readline is used by psql, a history file is automatically used.
This patch adds the special file name 'none', which if set as the
HISTFILE parameter, will cause psql not to use the history file.

I think it would be cleaner to use a separate \set variable to control
whether a history file is written, rather than needlessly overloading
the meaning of HISTFILE.

Why is this useful at all? There's already the -n (don't use readline)
switch.

Seems he wants readline without history, perhaps for security. Doesn't
setting HISTFILE to /dev/null work?

--
Bruce Momjian bruce@momjian.us
EnterpriseDB http://www.enterprisedb.com

+ If your life is a hard drive, Christ can be your backup. +

On 8/21/06, Bruce Momjian <bruce@momjian.us> wrote:

Tom Lane wrote:

Neil Conway <neilc@samurai.com> writes:

On Mon, 2006-08-21 at 19:27 +0300, Martin Atukunda wrote:

If readline is used by psql, a history file is automatically used.
This patch adds the special file name 'none', which if set as the
HISTFILE parameter, will cause psql not to use the history file.

I think it would be cleaner to use a separate \set variable to control
whether a history file is written, rather than needlessly overloading
the meaning of HISTFILE.

Why is this useful at all? There's already the -n (don't use readline)
switch.

Seems he wants readline without history, perhaps for security. Doesn't
setting HISTFILE to /dev/null work?

hmm, setting HISTFILE to /dev/null doesn't work on my MacOSX here. so
I whipped up this patch.

- Martin -

Am Freitag, 25. August 2006 17:03 schrieb Martin Atukunda:

hmm, setting HISTFILE to /dev/null doesn't work on my MacOSX here.

Please elaborate on "doesn't work".

--
Peter Eisentraut
http://developer.postgresql.org/~petere/

On 8/25/06, Peter Eisentraut <peter_e@gmx.net> wrote:

Am Freitag, 25. August 2006 17:03 schrieb Martin Atukunda:

hmm, setting HISTFILE to /dev/null doesn't work on my MacOSX here.

Please elaborate on "doesn't work".

without any .psqlrc file I get the following error when quitting a psql session:

could not save history to file "/Users/matlads/.psql_history": Invalid argument

When I set HISTFILE to /dev/null I get the following:

could not save history to file "/dev/null": Operation not permitted

- Martin -

"Martin Atukunda" <matlads@gmail.com> writes:

On 8/25/06, Peter Eisentraut <peter_e@gmx.net> wrote:

Please elaborate on "doesn't work".

without any .psqlrc file I get the following error when quitting a psql session:
could not save history to file "/Users/matlads/.psql_history": Invalid argument

That is fixed in CVS HEAD. The current coding looks like:

/*
* return value of write_history is not standardized across GNU
* readline and libedit. Therefore, check for errno becoming set
* to see if the write failed.
*/
errno = 0;
(void) write_history(fname);
if (errno == 0)
return true;

psql_error("could not save history to file \"%s\": %s\n",
fname, strerror(errno));

When I set HISTFILE to /dev/null I get the following:
could not save history to file "/dev/null": Operation not permitted

Hm. ktrace shows this happening:

23279 psql CALL open(0x302d70,0x601,0x1b6)
23279 psql NAMI "/dev/null"
23279 psql RET open 3
23279 psql CALL fchmod(0x3,0x180)
23279 psql RET fchmod -1 errno 1 Operation not permitted
23279 psql CALL close(0x3)
23279 psql RET close 0
23279 psql CALL write(0x2,0xbffff180,0x44)
23279 psql GIO fd 2 wrote 68 bytes
"could not save history to file "/dev/null": Operation not permitted
"
23279 psql RET write 68/0x44
23279 psql CALL exit(0)

There's probably no way to get Apple's libedit to not try the fchmod,
so what do we want to do here? Maybe special-case the string
"/dev/null"?

regards, tom lane

On 8/25/06, Tom Lane <tgl@sss.pgh.pa.us> wrote:

When I set HISTFILE to /dev/null I get the following:
could not save history to file "/dev/null": Operation not permitted

Hm. ktrace shows this happening:

23279 psql CALL open(0x302d70,0x601,0x1b6)
23279 psql NAMI "/dev/null"
23279 psql RET open 3
23279 psql CALL fchmod(0x3,0x180)
23279 psql RET fchmod -1 errno 1 Operation not permitted
23279 psql CALL close(0x3)
23279 psql RET close 0
23279 psql CALL write(0x2,0xbffff180,0x44)
23279 psql GIO fd 2 wrote 68 bytes
"could not save history to file "/dev/null": Operation not permitted
"
23279 psql RET write 68/0x44
23279 psql CALL exit(0)

There's probably no way to get Apple's libedit to not try the fchmod,
so what do we want to do here? Maybe special-case the string
"/dev/null"?

If this is OK, I can up with a patch that special cases /dev/null as a
HISTFILE if libedit is found.

- Martin -

"Martin Atukunda" <matlads@gmail.com> writes:

On 8/25/06, Tom Lane <tgl@sss.pgh.pa.us> wrote:

There's probably no way to get Apple's libedit to not try the fchmod,
so what do we want to do here? Maybe special-case the string
"/dev/null"?

If this is OK, I can up with a patch that special cases /dev/null as a
HISTFILE if libedit is found.

I was thinking of basically a one-liner addition to write_history
to skip the whole thing if strcmp(fname, DEVNULL) == 0. Should be
reasonably inoffensive on anyone's machine.

regards, tom lane

On 8/25/06, Tom Lane <tgl@sss.pgh.pa.us> wrote:

"Martin Atukunda" <matlads@gmail.com> writes:

On 8/25/06, Tom Lane <tgl@sss.pgh.pa.us> wrote:

There's probably no way to get Apple's libedit to not try the fchmod,
so what do we want to do here? Maybe special-case the string
"/dev/null"?

If this is OK, I can up with a patch that special cases /dev/null as a
HISTFILE if libedit is found.

I was thinking of basically a one-liner addition to write_history
to skip the whole thing if strcmp(fname, DEVNULL) == 0. Should be
reasonably inoffensive on anyone's machine.

I guess you meant saveHistory instead of write_history here. :)

something like the attached diff

- Martin -