Postgres Pro
Downloads
Home   >   mailing lists

Thread: BUG #1741: %i missing in snprintf implementation

BUG #1741: %i missing in snprintf implementation

From
"Tim Mauch"
Date:
The following bug has been logged online:

Bug reference:      1741
Logged by:          Tim Mauch
Email address:      tgmauch@yahoo.com
PostgreSQL version: 8.0.1 - 8.0.3
Operating system:   Redhat linux 8.0
Description:        %i missing in snprintf implementation
Details:

The implementation of snprintf/vsnprintf in src/port/snprintf.c does not
allow the use of %i ( %d equivalent ).  This likely isn't a problem for
postgres itself, but when a program compiles with -lpq and the standard c
libraries are not previously included the use of %i in the format string of
snprintf causes segmentation faults.

This is easily fixed by adding "case 'i':" to the same block of code handled
by "case 'd':" and case 'D':" in the implementation.

Re: BUG #1741: %i missing in snprintf implementation

From
Tom Lane
Date:
"Tim Mauch" <tgmauch@yahoo.com> writes:
> The implementation of snprintf/vsnprintf in src/port/snprintf.c does not
> allow the use of %i ( %d equivalent ).  This likely isn't a problem for
> postgres itself, but when a program compiles with -lpq and the standard c
> libraries are not previously included the use of %i in the format string of
> snprintf causes segmentation faults.
> This is easily fixed by adding "case 'i':" to the same block of code handled
> by "case 'd':" and case 'D':" in the implementation.

We are certainly not going to buy into the assumption that our snprintf
has to support every odd feature that anyone anywhere thinks snprintf
should have :-(.  The correct response to this is to figure out how your
program's calls got linked to our snprintf, and stop that from
happening.

AFAICS, our library should be defining the global symbol pg_snprintf not
snprintf.  Can you look into why that's not happening for you?

            regards, tom lane

Re: BUG #1741: %i missing in snprintf implementation

From
Bruce Momjian
Date:
Tom Lane wrote:
> "Tim Mauch" <tgmauch@yahoo.com> writes:
> > The implementation of snprintf/vsnprintf in src/port/snprintf.c does not
> > allow the use of %i ( %d equivalent ).  This likely isn't a problem for
> > postgres itself, but when a program compiles with -lpq and the standard c
> > libraries are not previously included the use of %i in the format string of
> > snprintf causes segmentation faults.
> > This is easily fixed by adding "case 'i':" to the same block of code handled
> > by "case 'd':" and case 'D':" in the implementation.
>
> We are certainly not going to buy into the assumption that our snprintf
> has to support every odd feature that anyone anywhere thinks snprintf
> should have :-(.  The correct response to this is to figure out how your
> program's calls got linked to our snprintf, and stop that from
> happening.
>
> AFAICS, our library should be defining the global symbol pg_snprintf not
> snprintf.  Can you look into why that's not happening for you?

Uh, the pg_snprintf macro fix is not in 8.0.X, only 8.1.  The good news,
however, is that very few platforms used our snprintf in 8.0 (the ones
that don't have 64-bit integer support in snprintf).

--
  Bruce Momjian                        |  http://candle.pha.pa.us
  pgman@candle.pha.pa.us               |  (610) 359-1001
  +  If your life is a hard drive,     |  13 Roberts Road
  +  Christ can be your backup.        |  Newtown Square, Pennsylvania 19073