Thread: Re: [COMMITTERS] pgsql: Add traceback information to PL/Python errors

Re: [COMMITTERS] pgsql: Add traceback information to PL/Python errors

From
Robert Haas
Date:
On Wed, Apr 6, 2011 at 3:37 PM, Peter Eisentraut <peter_e@gmx.net> wrote:
> Add traceback information to PL/Python errors
>
> This mimics the traceback information the Python interpreter prints
> with exceptions.
>
> Jan Urbański

On my system this spits out a warning:

plpython.c: In function ‘PLy_traceback’:
plpython.c:4487: warning: ‘s’ may be used uninitialized in this function
plpython.c:4487: note: ‘s’ was declared here

That appears to be a live bug, unless it's guaranteed that lineno will
always be > 0.

Also, the loop test should really be written as current < lineno,
rather than current != lineno, just in case we should manage to pass a
lineno < 0, which with the current code would go into the tank and
spin.

This part looks pretty sketchy, too:
   while (s && isspace((unsigned char) *s))       s++;

Perhaps we meant to test *s here.  It's hard to believe that we're
really intending to test whether the pointer has fallen off the end of
the address space and wrapped around to NULL.

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


Re: [COMMITTERS] pgsql: Add traceback information to PL/Python errors

From
Jan Urbański
Date:
On 07/04/11 23:01, Robert Haas wrote:
> On Wed, Apr 6, 2011 at 3:37 PM, Peter Eisentraut <peter_e@gmx.net> wrote:
>> Add traceback information to PL/Python errors
>>
>> This mimics the traceback information the Python interpreter prints
>> with exceptions.
>>
>> Jan Urbański
> 
> On my system this spits out a warning:
> 
> plpython.c: In function ‘PLy_traceback’:
> plpython.c:4487: warning: ‘s’ may be used uninitialized in this function
> plpython.c:4487: note: ‘s’ was declared here
> 
> That appears to be a live bug, unless it's guaranteed that lineno will
> always be > 0.

lineno should be > 0, unless Python is trying to tell us that the code
frame originates from before the function.

> Also, the loop test should really be written as current < lineno,
> rather than current != lineno, just in case we should manage to pass a
> lineno < 0, which with the current code would go into the tank and
> spin.

Yeah, good point.

> This part looks pretty sketchy, too:
> 
>     while (s && isspace((unsigned char) *s))
>         s++;
> 
> Perhaps we meant to test *s here.  It's hard to believe that we're
> really intending to test whether the pointer has fallen off the end of
> the address space and wrapped around to NULL.

Gah, so short a function and so many things that I managed to get wrong.

There's also this:
http://archives.postgresql.org/pgsql-hackers/2011-04/msg00334.php

Jan


Re: [COMMITTERS] pgsql: Add traceback information to PL/Python errors

From
Robert Haas
Date:
On Thu, Apr 7, 2011 at 5:06 PM, Jan Urbański <wulczer@wulczer.org> wrote:
> On 07/04/11 23:01, Robert Haas wrote:
>> On Wed, Apr 6, 2011 at 3:37 PM, Peter Eisentraut <peter_e@gmx.net> wrote:
>>> Add traceback information to PL/Python errors
>>>
>>> This mimics the traceback information the Python interpreter prints
>>> with exceptions.
>>>
>>> Jan Urbański
>>
>> On my system this spits out a warning:
>>
>> plpython.c: In function ‘PLy_traceback’:
>> plpython.c:4487: warning: ‘s’ may be used uninitialized in this function
>> plpython.c:4487: note: ‘s’ was declared here
>>
>> That appears to be a live bug, unless it's guaranteed that lineno will
>> always be > 0.
>
> lineno should be > 0, unless Python is trying to tell us that the code
> frame originates from before the function.
>
>> Also, the loop test should really be written as current < lineno,
>> rather than current != lineno, just in case we should manage to pass a
>> lineno < 0, which with the current code would go into the tank and
>> spin.
>
> Yeah, good point.
>
>> This part looks pretty sketchy, too:
>>
>>     while (s && isspace((unsigned char) *s))
>>         s++;
>>
>> Perhaps we meant to test *s here.  It's hard to believe that we're
>> really intending to test whether the pointer has fallen off the end of
>> the address space and wrapped around to NULL.
>
> Gah, so short a function and so many things that I managed to get wrong.

Patch?

> There's also this:
> http://archives.postgresql.org/pgsql-hackers/2011-04/msg00334.php

Yep.  I am assuming Peter will look at that one.

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


Re: [COMMITTERS] pgsql: Add traceback information to PL/Python errors

From
Jan Urbański
Date:
On 08/04/11 00:25, Robert Haas wrote:
> On Thu, Apr 7, 2011 at 5:06 PM, Jan Urbański <wulczer@wulczer.org> wrote:
>> On 07/04/11 23:01, Robert Haas wrote:
>>> On Wed, Apr 6, 2011 at 3:37 PM, Peter Eisentraut <peter_e@gmx.net> wrote:
>>>> Add traceback information to PL/Python errors
>>>>
>>>> This mimics the traceback information the Python interpreter prints
>>>> with exceptions.
>>>>
>>>> Jan Urbański
>>>
>>> On my system this spits out a warning:
>>>
>>> plpython.c: In function ‘PLy_traceback’:
>>> plpython.c:4487: warning: ‘s’ may be used uninitialized in this function
>>> plpython.c:4487: note: ‘s’ was declared here
>>>
>>> That appears to be a live bug, unless it's guaranteed that lineno will
>>> always be > 0.
>>
>> lineno should be > 0, unless Python is trying to tell us that the code
>> frame originates from before the function.
>>
>>> Also, the loop test should really be written as current < lineno,
>>> rather than current != lineno, just in case we should manage to pass a
>>> lineno < 0, which with the current code would go into the tank and
>>> spin.
>>
>> Yeah, good point.
>>
>>> This part looks pretty sketchy, too:
>>>
>>>     while (s && isspace((unsigned char) *s))
>>>         s++;
>>>
>>> Perhaps we meant to test *s here.  It's hard to believe that we're
>>> really intending to test whether the pointer has fallen off the end of
>>> the address space and wrapped around to NULL.
>>
>> Gah, so short a function and so many things that I managed to get wrong.
>
> Patch?

Attached.

>> There's also this:
>> http://archives.postgresql.org/pgsql-hackers/2011-04/msg00334.php
>
> Yep.  I am assuming Peter will look at that one.

I guess so. This only fixes the things you noticed.

Jan

Attachment