Re: Potential reference miscounts and segfaults in plpython.c

On Mon, Feb 20, 2012 at 5:05 AM, Jan Urbański <wulczer(at)wulczer(dot)org> wrote:
> On 20/02/12 04:29, Tom Lane wrote:
>> =?UTF-8?B?SmFuIFVyYmHFhHNraQ==?= <wulczer(at)wulczer(dot)org> writes:
>>>> On 18/02/12 21:17, Tom Lane wrote:
>>>>> Dave Malcolm at Red Hat has been working on a static code analysis tool
>>>>> for Python-related C code.  He reports here on some preliminary results
>>>>> for plpython.c:
>>>>> https://bugzilla.redhat.com/show_bug.cgi?id=795011
>>
>>> Here's a patch that fixes everything I was sure was an actual bug. The
>>> rest of the warnings seem to be caused by the tool not knowing that
>>> elog(ERROR) throws a longjmp and things like "we never unref this
>>> object, so it can't disappear mid-execution".
>>
>> My only comment is whether elog(ERROR) is appropriate, ie, do we consider
>> these to be internal errors that users will never see in practice?
>
> AFAICS these errors can only happen on out of memory conditions or other
> internal errors (like trying to create a list with a negative length).

We typically want out of memory to use ereport, so that it gets
translated. For example, in fd.c we have:

ereport(FATAL,
(errcode(ERRCODE_OUT_OF_MEMORY),
errmsg("out of memory")));

Trying to create a list with a negative length sounds similar.

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