Обсуждение: [HACKERS] Large objects

Поиск
Список
Период
Сортировка

[HACKERS] Large objects

От
Sean Lyndersay
Дата:
The ToDo list includes the following as an enhancement in version 6.1:
  Large objects
      overwriting blocks has problems
      there are other problems, too.
      Fix large object mapping scheme
      not to stuff everything as files in a single directory

Does this mean that large objects work now? If so, is the libpq++
testlo.cc example supposed to work?

thx,
sean


________________________________________________________________________
Sean Lyndersay                        Time is natures way of making sure
lynders@hcs.harvard.edu                 things don't happen all at once.
________________________________________________________________________
[finger for all other info]          http://www.hcs.harvard.edu/~lynders

------------------------------

Re: [HACKERS] Large objects

От
Bruce Momjian
Дата:
>
>
> The ToDo list includes the following as an enhancement in version 6.1:
>   Large objects
>       overwriting blocks has problems
>       there are other problems, too.
>       Fix large object mapping scheme
>       not to stuff everything as files in a single directory
>
> Does this mean that large objects work now? If so, is the libpq++
> testlo.cc example supposed to work?

As the maintainer of the todo list, I can say, "I don't know."

The TODO list is only a list of REPORTED problems.

- --
Bruce Momjian
maillist@candle.pha.pa.us

------------------------------

Re: [HACKERS] Large objects

От
Sean Lyndersay
Дата:
On Mon, 9 Jun 1997, Bruce Momjian wrote:

> >
> > The ToDo list includes the following as an enhancement in version 6.1:
> >   Large objects
> >       overwriting blocks has problems
> >       there are other problems, too.
> >       Fix large object mapping scheme
> >       not to stuff everything as files in a single directory
> >
> > Does this mean that large objects work now? If so, is the libpq++
> > testlo.cc example supposed to work?
>
> As the maintainer of the todo list, I can say, "I don't know."
>
> The TODO list is only a list of REPORTED problems.
>

Okay, how about this:
- -- copy start --
       CHANGES IN THE 6.1 RELEASE

 Bug Fixes
 ---------
[snip]
large object fixes(Sven)

- -- copy end --

guess we'll wait until sven reads and comments. :)

meanwhile, i'll see if i can find the problem. For those interested, the
test program failes when trying to read the large object out of the
database with the following message:
 "OTICE:LockRelease: locktable lookup failed, no lock"


sean

________________________________________________________________________
Sean Lyndersay                        Time is natures way of making sure
lynders@hcs.harvard.edu                 things don't happen all at once.
________________________________________________________________________
[finger for all other info]          http://www.hcs.harvard.edu/~lynders

------------------------------

Re: [HACKERS] Large objects

От
Raymond Toy
Дата:
>>>>> "Sean" == Sean Lyndersay <lynders@hcs.harvard.edu> writes:


    Sean> On Mon, 9 Jun 1997, Bruce Momjian wrote:
    >> >
    >> > The ToDo list includes the following as an enhancement in
    >> > version 6.1:
    >> >   Large objects
    >> >       overwriting blocks has problems there are other
    >> >       problems, too.  Fix large object mapping scheme not to
    >> >       stuff everything as files in a single directory
    >> >
    >> > Does this mean that large objects work now? If so, is the
    >> > libpq++ testlo.cc example supposed to work?
    >>
    >> As the maintainer of the todo list, I can say, "I don't know."
    >>
    >> The TODO list is only a list of REPORTED problems.
    >>

    Sean> Okay, how about this:
    Sean> -- copy start --
    Sean>        CHANGES IN THE 6.1 RELEASE

    Sean>  Bug Fixes ---------
    Sean> [snip] large object fixes(Sven)

    Sean> -- copy end --

    Sean> guess we'll wait until sven reads and comments. :)

I've been pounding on large objects for the last few days.  It's
pretty stable, much more so than earlier versions.  However, I've only
tested the lo_import, lo_export SQL functions, and the lo_read,
lo_write, and lo_create functions in libpq.  They seem to work well
now.

    Sean> meanwhile, i'll see if i can find the problem. For those
    Sean> interested, the test program failes when trying to read the
    Sean> large object out of the database with the following message:
    Sean>  "OTICE:LockRelease: locktable lookup failed, no lock"

I've seen this.  I think this means either you have nested begin/end
transactions or forgot a begin/end transaction block, but have not
checked into it yet.

Ray

------------------------------

Re: [HACKERS] Large objects

От
Bruce Momjian
Дата:
Shall we install Raymonds large object patches in 6.1?  Looks like he
has done some serious testing, so it may be a good idea.

>
> >>>>> "Sean" == Sean Lyndersay <lynders@hcs.harvard.edu> writes:
>
>
>     Sean> On Mon, 9 Jun 1997, Bruce Momjian wrote:
>     >> >
>     >> > The ToDo list includes the following as an enhancement in
>     >> > version 6.1:
>     >> >   Large objects
>     >> >       overwriting blocks has problems there are other
>     >> >       problems, too.  Fix large object mapping scheme not to
>     >> >       stuff everything as files in a single directory
>     >> >
>     >> > Does this mean that large objects work now? If so, is the
>     >> > libpq++ testlo.cc example supposed to work?
>     >>
>     >> As the maintainer of the todo list, I can say, "I don't know."
>     >>
>     >> The TODO list is only a list of REPORTED problems.
>     >>
>
>     Sean> Okay, how about this:
>     Sean> -- copy start --
>     Sean>        CHANGES IN THE 6.1 RELEASE
>
>     Sean>  Bug Fixes ---------
>     Sean> [snip] large object fixes(Sven)
>
>     Sean> -- copy end --
>
>     Sean> guess we'll wait until sven reads and comments. :)
>
> I've been pounding on large objects for the last few days.  It's
> pretty stable, much more so than earlier versions.  However, I've only
> tested the lo_import, lo_export SQL functions, and the lo_read,
> lo_write, and lo_create functions in libpq.  They seem to work well
> now.
>
>     Sean> meanwhile, i'll see if i can find the problem. For those
>     Sean> interested, the test program failes when trying to read the
>     Sean> large object out of the database with the following message:
>     Sean>  "OTICE:LockRelease: locktable lookup failed, no lock"
>
> I've seen this.  I think this means either you have nested begin/end
> transactions or forgot a begin/end transaction block, but have not
> checked into it yet.
>
> Ray
>
>
>
>


- --
Bruce Momjian
maillist@candle.pha.pa.us

------------------------------

End of hackers-digest V1 #381
*****************************

Re: [HACKERS] Large objects

От
Sean Lyndersay
Дата:
> >
> > I've been pounding on large objects for the last few days.  It's
> > pretty stable, much more so than earlier versions.  However, I've only
> > tested the lo_import, lo_export SQL functions, and the lo_read,
> > lo_write, and lo_create functions in libpq.  They seem to work well
> > now.
> >
> >     Sean> meanwhile, i'll see if i can find the problem. For those
> >     Sean> interested, the test program failes when trying to read the
> >     Sean> large object out of the database with the following message:
> >     Sean>  "OTICE:LockRelease: locktable lookup failed, no lock"
> >
> > I've seen this.  I think this means either you have nested begin/end
> > transactions or forgot a begin/end transaction block, but have not
> > checked into it yet.
> >

Can't be that. I'm using the testlo.cc program in the libpq++/examples
directory. It has a begin and an end command. For reference, I'm using the
snapshot day before yesterday.

My preference would be to have these large object patches asap. It seems
that they are broken right now. Either that or libpq++ is broken. If
Raymond is having no problems with libpq, then I can debug libpq++ and see
what's happening. I'd rather not repeat anything he's done.

thanks
sean

________________________________________________________________________
Sean Lyndersay                        Time is natures way of making sure
lynders@hcs.harvard.edu                 things don't happen all at once.
________________________________________________________________________
[finger for all other info]          http://www.hcs.harvard.edu/~lynders

------------------------------

Re: [HACKERS] Large objects

От
Raymond Toy
Дата:
>>>>> "Sean" == Sean Lyndersay <lynders@hcs.harvard.edu> writes:


    Sean> Can't be that. I'm using the testlo.cc program in the
    Sean> libpq++/examples directory. It has a begin and an end
    Sean> command. For reference, I'm using the snapshot day before
    Sean> yesterday.

    Sean> My preference would be to have these large object patches
    Sean> asap. It seems that they are broken right now. Either that
    Sean> or libpq++ is broken. If Raymond is having no problems with
    Sean> libpq, then I can debug libpq++ and see what's
    Sean> happening. I'd rather not repeat anything he's done.

Ok.  Just a thought.  I have not tried libpq++, and all of my testing
has been either with psql or my homegrown STk to PostgreSQL
interface (pgstk).

If you have the version from two or three days ago, you are working
with the same version as I am (was) using.  You are only missing the
patch I posted on Sunday which fixes a bug in the backend for
lo_import and lo_export.  The sample code in test/examples/testlo.c
works just fine for me (but needed to include <fcntl.h>) so I suspect
either a simple bug in either testlo.cc or libpq++.

Ray

------------------------------