Обсуждение: not checking value returned from palloc?

Looking through the postgresql source code, I notice that there are many places 
were palloc is used but the return value is not checked to see if it is null. 
There are a few places such as:
        if (!PointerIsValid(result = palloc(CASH_BUFSZ + 2 - count +    strlen(nsymbol))))            ereport(ERROR,
               (errcode(ERRCODE_OUT_OF_MEMORY),                     errmsg("out of memory")));
 

(taken from src/backend/utils/adt/cash.c), but at least within the that same 
directory most occurrences of palloc are not checked.

Is this sloppy programming, or is there an automagical thing going on with 
#defines that I'm just not seeing?

If it is just sloppy, perhaps we could use a new define in palloc.h, such as:

#define palloc_or_die(ptr,sz) \do { \    ptr = palloc(sz); \    if (!ptr) \    { \        ereport(ERROR, \
(errcode(ERRCODE_OUT_OF_MEMORY),\             errmsg("Out of memory"))); \    } \} while(0);    
 
And then, in all places where the code does not currently check the return value 
of palloc, the code could be changed to use palloc_or_die instead.  Of course, 
I'd be happy if someone has a better name for the macro, perhaps something more 
brief?

I can go over the code "with a fine tooth comb" and replace the offending 
occurrences.  Does the community think this is a good idea?

mark

Mark Dilger wrote:
> Looking through the postgresql source code, I notice that there are
> many places were palloc is used but the return value is not checked
> to see if it is null.

palloc will throw an exception if it cannot fulfill the request.  Code 
that checks the return value for null is in fact a waste.

-- 
Peter Eisentraut
http://developer.postgresql.org/~petere/

Mark Dilger <pgsql@markdilger.com> writes:
> Looking through the postgresql source code, I notice that there are many places 
> were palloc is used but the return value is not checked to see if it is null. 
Any place that *does* check it is incorrect (in the sense of being
wasted code space, not functionally incorrect).  Please read the memmgr
documentation.
        regards, tom lane

Peter Eisentraut wrote:
> Mark Dilger wrote:
> 
>>Looking through the postgresql source code, I notice that there are
>>many places were palloc is used but the return value is not checked
>>to see if it is null.
> 
> 
> palloc will throw an exception if it cannot fulfill the request.  Code 
> that checks the return value for null is in fact a waste.
> 

Interesting.  So the patch should go the other way, and remove the checks that 
are currently in the code?

Mark Dilger wrote:
> Peter Eisentraut wrote:
> >Mark Dilger wrote:
> >
> >>Looking through the postgresql source code, I notice that there are
> >>many places were palloc is used but the return value is not checked
> >>to see if it is null.
> >
> >palloc will throw an exception if it cannot fulfill the request.  Code 
> >that checks the return value for null is in fact a waste.
> 
> Interesting.  So the patch should go the other way, and remove the checks 
> that are currently in the code?

Yes, if you find any such place please submit a patch.

-- 
Alvaro Herrera                                http://www.CommandPrompt.com/
PostgreSQL Replication, Consulting, Custom Development, 24x7 support

Alvaro Herrera wrote:
> Mark Dilger wrote:
> > Peter Eisentraut wrote:
> > >Mark Dilger wrote:
> > >
> > >>Looking through the postgresql source code, I notice that there are
> > >>many places were palloc is used but the return value is not checked
> > >>to see if it is null.
> > >
> > >palloc will throw an exception if it cannot fulfill the request.  Code 
> > >that checks the return value for null is in fact a waste.
> > 
> > Interesting.  So the patch should go the other way, and remove the checks 
> > that are currently in the code?
> 
> Yes, if you find any such place please submit a patch.

I see one in cash.c that I will remove.

--  Bruce Momjian   http://candle.pha.pa.us SRA OSS, Inc.   http://www.sraoss.com
 + If your life is a hard drive, Christ can be your backup. +

On Sun, 2006-03-19 at 17:14 -0500, Bruce Momjian wrote:
> I see one in cash.c that I will remove.

I've already checked in a fix for that, as well as a few other places
that made similar mistakes -- sorry for stepping on your toes.

-Neil