Re: installer problems

Hi,

I'm having problems installing the database, see:

anfitrite:/opt# ./postgresql-9.0.0-1-linux.bin

Error: Error running /tmp/postgresql_installer/getlocales : child killed:
floating-point exception
Press [Enter] to continue :
anfitrite:/opt#

anfitrite:/opt# uname -a
Linux anfitrite 2.6.18-6-686 #1 SMP Fri Feb 19 23:40:03 UTC 2010 i686
GNU/Linux

anfitrite:/opt# cat /etc/debian_version
4.0

Thanks.
Samuel
rettore(at)gmail(dot)com

On Thu, Sep 23, 2010 at 3:19 PM, Samuel Rettore <rettore(at)gmail(dot)com> wrote:
> Hi,
>
> I'm having problems installing the database, see:
>
> anfitrite:/opt# ./postgresql-9.0.0-1-linux.bin
>
> Error: Error running /tmp/postgresql_installer/getlocales  : child killed:
> floating-point exception
> Press [Enter] to continue :
> anfitrite:/opt#
>
>
> anfitrite:/opt# uname -a
> Linux anfitrite 2.6.18-6-686 #1 SMP Fri Feb 19 23:40:03 UTC 2010 i686
> GNU/Linux
>
> anfitrite:/opt# cat /etc/debian_version
> 4.0

Debian is not a supported platform for the installers - we don't see
enough demand for it to justify the testing requirements I'm afraid.

--
Dave Page
Blog: http://pgsnake.blogspot.com
Twitter: @pgsnake

EnterpriseDB UK: http://www.enterprisedb.com
The Enterprise Postgres Company

On tor, 2010-09-23 at 16:45 +0100, Dave Page wrote:
> On Thu, Sep 23, 2010 at 3:19 PM, Samuel Rettore <rettore(at)gmail(dot)com> wrote:
> > Hi,
> >
> > I'm having problems installing the database, see:
> >
> > anfitrite:/opt# ./postgresql-9.0.0-1-linux.bin
> >
> > Error: Error running /tmp/postgresql_installer/getlocales : child killed:
> > floating-point exception
> > Press [Enter] to continue :
> > anfitrite:/opt#
> >
> >
> > anfitrite:/opt# uname -a
> > Linux anfitrite 2.6.18-6-686 #1 SMP Fri Feb 19 23:40:03 UTC 2010 i686
> > GNU/Linux
> >
> > anfitrite:/opt# cat /etc/debian_version
> > 4.0
>
> Debian is not a supported platform for the installers - we don't see
> enough demand for it to justify the testing requirements I'm afraid.

Surely the error message can be polished, though?

On Thu, Sep 23, 2010 at 11:46 PM, Peter Eisentraut <peter_e(at)gmx(dot)net> wrote:
> On tor, 2010-09-23 at 16:45 +0100, Dave Page wrote:
>> On Thu, Sep 23, 2010 at 3:19 PM, Samuel Rettore <rettore(at)gmail(dot)com> wrote:
>> > Hi,
>> >
>> > I'm having problems installing the database, see:
>> >
>> > anfitrite:/opt# ./postgresql-9.0.0-1-linux.bin
>> >
>> > Error: Error running /tmp/postgresql_installer/getlocales  : child killed:
>> > floating-point exception
>> > Press [Enter] to continue :
>> > anfitrite:/opt#
>> >
>> >
>> > anfitrite:/opt# uname -a
>> > Linux anfitrite 2.6.18-6-686 #1 SMP Fri Feb 19 23:40:03 UTC 2010 i686
>> > GNU/Linux
>> >
>> > anfitrite:/opt# cat /etc/debian_version
>> > 4.0
>>
>> Debian is not a supported platform for the installers - we don't see
>> enough demand for it to justify the testing requirements I'm afraid.
>
> Surely the error message can be polished, though?

Well we don't know what errors we might get on unsupported platforms
until we test them, and we certainly don't have the resources to test
every unsupported platform. It's hard enough to test the 40 or so we
do support.

--
Dave Page
Blog: http://pgsnake.blogspot.com
Twitter: @pgsnake

EnterpriseDB UK: http://www.enterprisedb.com
The Enterprise Postgres Company

On tor, 2010-09-23 at 23:56 +0100, Dave Page wrote:
> > Surely the error message can be polished, though?
>
> Well we don't know what errors we might get on unsupported platforms
> until we test them, and we certainly don't have the resources to test
> every unsupported platform. It's hard enough to test the 40 or so we
> do support.

I don't accept that argument, but FWIW, it works just fine on my Debian
installation, so nothing to worry about.

On Fri, Sep 24, 2010 at 6:28 AM, Peter Eisentraut <peter_e(at)gmx(dot)net> wrote:
> On tor, 2010-09-23 at 23:56 +0100, Dave Page wrote:
>> > Surely the error message can be polished, though?
>>
>> Well we don't know what errors we might get on unsupported platforms
>> until we test them, and we certainly don't have the resources to test
>> every unsupported platform. It's hard enough to test the 40 or so we
>> do support.
>
> I don't accept that argument,

How else would you suggest we find out what may go wrong on an
unsupported platform?

--
Dave Page
Blog: http://pgsnake.blogspot.com
Twitter: @pgsnake

EnterpriseDB UK: http://www.enterprisedb.com
The Enterprise Postgres Company

Dave Page <dpage(at)pgadmin(dot)org> writes:
> On Fri, Sep 24, 2010 at 6:28 AM, Peter Eisentraut <peter_e(at)gmx(dot)net> wrote:
>> On tor, 2010-09-23 at 23:56 +0100, Dave Page wrote:
>>> > Surely the error message can be polished, though?
>
> How else would you suggest we find out what may go wrong on an
> unsupported platform?

My guess here is that you could use `lsb_release -a` or some other
system dependant facilities to know the current platform, then error out
with a nice error message if it's not one you know you support.

This way the installer won't even try to run and won't go wrong on an
unsupported platform…

Regards,
--
dim

On Fri, Sep 24, 2010 at 11:30 AM, Dimitri Fontaine
<dfontaine(at)hi-media(dot)com> wrote:
> Dave Page <dpage(at)pgadmin(dot)org> writes:
>> On Fri, Sep 24, 2010 at 6:28 AM, Peter Eisentraut <peter_e(at)gmx(dot)net> wrote:
>>> On tor, 2010-09-23 at 23:56 +0100, Dave Page wrote:
>>>> > Surely the error message can be polished, though?
>>
>> How else would you suggest we find out what may go wrong on an
>> unsupported platform?
>
> My guess here is that you could use `lsb_release -a` or some other
> system dependant facilities to know the current platform, then error out
> with a nice error message if it's not one you know you support.
>
> This way the installer won't even try to run and won't go wrong on an
> unsupported platform…

I didn't know that existed - thanks. Unfortunately the first box I
tried it on didn't have it, but that's quite old and not a supported
platform anyway. if it does exist on everything we support, then that
could work.

--
Dave Page
Blog: http://pgsnake.blogspot.com
Twitter: @pgsnake

EnterpriseDB UK: http://www.enterprisedb.com
The Enterprise Postgres Company

On Thu, Sep 23, 2010 at 11:45 AM, Dave Page <dpage(at)pgadmin(dot)org> wrote:
> On Thu, Sep 23, 2010 at 3:19 PM, Samuel Rettore <rettore(at)gmail(dot)com> wrote:
>> Hi,
>>
>> I'm having problems installing the database, see:
>>
>> anfitrite:/opt# ./postgresql-9.0.0-1-linux.bin
>>
>> Error: Error running /tmp/postgresql_installer/getlocales  : child killed:
>> floating-point exception
>> Press [Enter] to continue :
>> anfitrite:/opt#
>>
>>
>> anfitrite:/opt# uname -a
>> Linux anfitrite 2.6.18-6-686 #1 SMP Fri Feb 19 23:40:03 UTC 2010 i686
>> GNU/Linux
>>
>> anfitrite:/opt# cat /etc/debian_version
>> 4.0
>
> Debian is not a supported platform for the installers - we don't see
> enough demand for it to justify the testing requirements I'm afraid.

So I think what that means for the OP is that he should install
PostgreSQL by some other method - for example, apt-get, or compiling
from source.

--
Robert Haas
EnterpriseDB: http://www.enterprisedb.com
The Enterprise Postgres Company

On fre, 2010-09-24 at 09:01 +0100, Dave Page wrote:
> How else would you suggest we find out what may go wrong on an
> unsupported platform?

Well, some amount of robust and defensive programming can probably
create better error messages in even the most unknown environment.
Especially since the product you're creating is to some degree intended
for newbies.

Now I suspect that the problem of the user might be unrelated to the
installer not actually supporting the host platform, but I'm questioning
the approach of dismissing error messages like "child killed" and
"floating-point exception" with "unsupported, don't care".

Note also that actually finding out which platforms are supported
requires the unsuspecting user to download a PDF file and scroll down a
few pages, whereas the way it's presented for download might give the
impression in runs everywhere.

On fre, 2010-09-24 at 11:38 +0100, Dave Page wrote:
> I didn't know that existed - thanks. Unfortunately the first box I
> tried it on didn't have it, but that's quite old and not a supported
> platform anyway. if it does exist on everything we support, then that
> could work.

It's tempting, but I would caution against that. The installer might
work just fine (or 95% fine) on unsupported platforms, but if you just
shut people out you won't know. Of course you might want to print a
*warning* based on lsb_release.

Excerpts from Peter Eisentraut's message of vie sep 24 10:55:58 -0400 2010:
> On fre, 2010-09-24 at 11:38 +0100, Dave Page wrote:
> > I didn't know that existed - thanks. Unfortunately the first box I
> > tried it on didn't have it, but that's quite old and not a supported
> > platform anyway. if it does exist on everything we support, then that
> > could work.
>
> It's tempting, but I would caution against that. The installer might
> work just fine (or 95% fine) on unsupported platforms, but if you just
> shut people out you won't know. Of course you might want to print a
> *warning* based on lsb_release.

I think the logic is if lsb_release exists on all supported platforms,
then use it to determine whether this is a supported platform or not.
If it doesn't exist, then you know it's not a supported platform in the
first place.

--
Álvaro Herrera <alvherre(at)commandprompt(dot)com>
The PostgreSQL Company - Command Prompt, Inc.
PostgreSQL Replication, Consulting, Custom Development, 24x7 support

On Fri, Sep 24, 2010 at 3:52 PM, Peter Eisentraut <peter_e(at)gmx(dot)net> wrote:
> On fre, 2010-09-24 at 09:01 +0100, Dave Page wrote:
>> How else would you suggest we find out what may go wrong on an
>> unsupported platform?
>
> Well, some amount of robust and defensive programming can probably
> create better error messages in even the most unknown environment.
> Especially since the product you're creating is to some degree intended
> for newbies.

We're pretty restricted on how we can handle errors in the installer.
When we want to do more complex things, we call out to helper programs
and scripts where we can more effectively handle errors, but when it's
one of those that is failing to even run properly it's difficult to
deal with every possibility.

> Now I suspect that the problem of the user might be unrelated to the
> installer not actually supporting the host platform, but I'm questioning
> the approach of dismissing error messages like "child killed" and
> "floating-point exception" with "unsupported, don't care".

Well I'm struggling to see where floating-point exception might come
from anyway. The code that apparently gave the error doesn't have any
floating point arithmetic in it, so it must be coming from some other
library or glibc. Much as it would be nice to know whats going on,
it's simply not practical to chase such bugs on platforms we haven't
tested so "unsupported" really is the only option.

> Note also that actually finding out which platforms are supported
> requires the unsuspecting user to download a PDF file and scroll down a
> few pages, whereas the way it's presented for download might give the
> impression in runs everywhere.

The download page states exactly what distros we consider supported.

--
Dave Page
Blog: http://pgsnake.blogspot.com
Twitter: @pgsnake

EnterpriseDB UK: http://www.enterprisedb.com
The Enterprise Postgres Company

On Fri, Sep 24, 2010 at 4:09 PM, Alvaro Herrera
<alvherre(at)commandprompt(dot)com> wrote:
> Excerpts from Peter Eisentraut's message of vie sep 24 10:55:58 -0400 2010:
>> On fre, 2010-09-24 at 11:38 +0100, Dave Page wrote:
>> > I didn't know that existed - thanks. Unfortunately the first box I
>> > tried it on didn't have it, but that's quite old and not a supported
>> > platform anyway. if it does exist on everything we support, then that
>> > could work.
>>
>> It's tempting, but I would caution against that.  The installer might
>> work just fine (or 95% fine) on unsupported platforms, but if you just
>> shut people out you won't know.  Of course you might want to print a
>> *warning* based on lsb_release.
>
> I think the logic is if lsb_release exists on all supported platforms,
> then use it to determine whether this is a supported platform or not.
> If it doesn't exist, then you know it's not a supported platform in the
> first place.

Yeah. We wouldn't prevent execution on an unsupported platform - just
warn the user that it hasn't been tested and therefore might not work
so well.

--
Dave Page
Blog: http://pgsnake.blogspot.com
Twitter: @pgsnake

EnterpriseDB UK: http://www.enterprisedb.com
The Enterprise Postgres Company

On fre, 2010-09-24 at 16:10 +0100, Dave Page wrote:
> The download page states exactly what distros we consider supported.

Sorry, I was actually looking at the Postgres Plus download page.