Re: BUG #14287: psql_history gets wiped out

Поиск
Список
Период
Сортировка
От Rick Otten
Тема Re: BUG #14287: psql_history gets wiped out
Дата
Msg-id 59009b1730d8b42e4d08226c3fe08058@www.windfish.net
обсуждение исходный текст
Ответ на Re: BUG #14287: psql_history gets wiped out  ("Daniel Verite" <daniel@manitou-mail.org>)
Список pgsql-bugs

Aha.  Indeed my path was going straight to the psql binary without passing through the /usr/bin/psql perl wrapper.

Even when I use the wrapper, I was still able to reproduce wiping out my .psql_history file by closing the window before exiting psql.

 

On 2016-08-12 12:21, Daniel Verite wrote:

	Rick Otten wrote:
think I have readline installed on the system: $ dpkg -l | grep readline ii libreadline5:amd64 5.2+dfsg-2 amd64 GNU readline and history libraries, run-time libraries ii libreadline6:amd64 6.3-4ubuntu2 amd64 GNU readline and history libraries, run-time libraries ii readline-common
According to the strace output, the replacement of libedit
by libreadline at runtime does not happen in your case
(the "_HiStOrY_V2_" string is from libedit).

This replacement is done by pg_wrapper, a Perl script to which
/usr/bin/psql is normally linked:
$ ls -l /usr/bin/psql
/usr/bin/psql -> ../share/postgresql-common/pg_wrapper
The script pushes libreadline over libedit through LD_PRELOAD, and
launches the actual binary located in 
/usr/lib/postgresql/X.Y/bin/psql (where X.Y is the current cluster's
PG version).

You may want to check if that script is not entirely sidestepped in
your setup, as that would explain why libedit gets used. For instance,
a modified PATH or psql being a shell alias could lead to that.

Aside from the bug resulting in an empty file, libedit
will ignore readline's history and remove it from the file
when saving, which is annoying by itself if using it by accident.

Best regards,

 

В списке pgsql-bugs по дате отправления:

Предыдущее
От: "Daniel Verite"
Дата:
Сообщение: Re: BUG #14287: psql_history gets wiped out
Следующее
От: "VJe@n"
Дата:
Сообщение: PgAdmin3 bug on application_name