Problem V8.1.4 - providing pwd for commandline tools doesn't work anymore
The latest version (V8.1.4) breaks the setup process of our
application's installation wizard. We used to call "createdb.exe" and
piped the password for the postgres user (which has been entered by the
user in our setup wizard's dialogs) into it.
With version V8.1.4 this is not possible anymore, the commandline
process prompts for the password nevertheless! This causes our setup to
"hang" (of course it doesn't hang, it waits for the invisible
commandline process to return, which waits for a user input that will
never complete as the user cannot and shall not see these commandline
tools).
Any suggestion? (V8.1.3 and earlier did work fine in this respect.)
Thank you in advance and sorry if this is a (new) V8.1.4 FAQ, I didn't
found anything in this respect.
Best Regards,
Alexander.
The latest version (V8.1.4) breaks the setup process of our
application's installation wizard. We used to call
"createdb.exe" and piped the password for the postgres user
(which has been entered by the user in our setup wizard's
dialogs) into it.With version V8.1.4 this is not possible anymore, the
commandline process prompts for the password nevertheless!
This causes our setup to "hang" (of course it doesn't hang,
it waits for the invisible commandline process to return,
which waits for a user input that will never complete as the
user cannot and shall not see these commandline tools).Any suggestion? (V8.1.3 and earlier did work fine in this respect.)
Actually, no, v8.1.3 and earlier are the ones that are broken in this
respect, and it was fixed in 8.1.4.
You need to put the password either in a pgpass.conf file
(http://www.postgresql.org/docs/8.1/static/libpq-pgpass.html), or put it
in the PGPASSWORD environment variable
(http://www.postgresql.org/docs/8.1/static/libpq-envars.html - it's
secure on Windows per the comment)
//Magnus
Import Notes
Resolved by subject fallback
I don't understand why the change should be "more secure", but I see
that nobody took care about the possible consequences for installation
scripts, third party applications and so on. :-(((
If you change such essential behaviour (the old behaviour of the
commandline tools did work well since 8.0!) and cause such an
incompatibility, you should have at least clearly stated this in the
update readme. The comment "Fix problem with password prompting on some
Win32 systems" is a little bit "short" regarding such an immens impact
on any installation script...
Sigh,
Alexander.
Show quoted text
on 31.05.2006 10:00 Magnus Hagander wrote:
The latest version (V8.1.4) breaks the setup process of our
application's installation wizard. We used to call
"createdb.exe" and piped the password for the postgres user
(which has been entered by the user in our setup wizard's
dialogs) into it.With version V8.1.4 this is not possible anymore, the
commandline process prompts for the password nevertheless!
This causes our setup to "hang" (of course it doesn't hang,
it waits for the invisible commandline process to return,
which waits for a user input that will never complete as the
user cannot and shall not see these commandline tools).Any suggestion? (V8.1.3 and earlier did work fine in this respect.)
Actually, no, v8.1.3 and earlier are the ones that are broken in this
respect, and it was fixed in 8.1.4.You need to put the password either in a pgpass.conf file
(http://www.postgresql.org/docs/8.1/static/libpq-pgpass.html), or put it
in the PGPASSWORD environment variable
(http://www.postgresql.org/docs/8.1/static/libpq-envars.html - it's
secure on Windows per the comment)//Magnus
---------------------------(end of broadcast)---------------------------
TIP 5: don't forget to increase your free space map settings
I don't understand why the change should be "more secure",
but I see that nobody took care about the possible
consequences for installation scripts, third party
applications and so on. :-(((
If the installation scripts and third party apps followed the standard
way of doing it, they would not be affected. What has been done is to
change the win32 version of the tools to behave the exact same way as
the tools have behaved on Unix for years.
Previously, the win32 implementation was not in sync with the
documentation. After 8.1.4, it now is.
(There's also the fact that the old, incorrect, handling had a serious
bug in it that could make it impossible to connect to the database if
you had a c:\dev directory on your machine)
If you change such essential behaviour (the old behaviour of
the commandline tools did work well since 8.0!) and cause
such an incompatibility, you should have at least clearly
stated this in the update readme. The comment "Fix problem
with password prompting on some
Win32 systems" is a little bit "short" regarding such an
immens impact on any installation script...
That, however, I agree with. It should've been clearer in the release
notes. We just didn't anticipate that people relied on teh old buggy
behaviour.
Perhaps that's worth updating in the release notes before we release
8.1.5, what do people think?
//Magnus
Import Notes
Resolved by subject fallback