Try this:

cat >test_4096.c <<EOF
#include <stdio.h>

int main()
{
         unsigned long long a;

         a = -4096;

         printf("%llu\n",a);

         return 0;
}
EOF
cc -g -Wall test_4096.c -o test_4096
./test_4096

-4096 is represented as 0xfffffffffffff000 (as a signed 64 bit 
integer). If you cast it to an unsigned 64 bit integer, you get 
18446744073709547520.

The probability that something calls malloc requesting 18 billion 
gigabytes is low. The probability that something calls malloc 
requesting a negative size is much higher.

You might want to try strace'ing a one-liner that does a "malloc(-1)" 
(or more certainly "malloc(-4097)" depending on how the "rounding" is 
done?) just for fun :-)

cat >test_malloc.c <<EOF
#include <stdlib.h>

int main()
{
         malloc(-1);
         return 0;
}
EOF
cc -g -Wall test_malloc.c -o test_malloc
strace ./test_malloc

Results differ from platform to platform, depending mainly on the 
specific implementation of malloc. FreeBSD doesn't yield any "fun" 
results, but apparently your OS does.

I have gone quickly through the slonik source, including the flex and 
bison generated code, and I don't offhand see any reason for a 
negative malloc to happen, but there must be a case where something 
ends up requesting a negative number of bytes. My guess was around 
the yyleng manipulation, but it seems (?) to be correct, so it must 
be somewhere else, or it is your version of flex or bison that is buggy.

What are your versions of flex and bison?

Jacques.

At 21:53 11/06/2008, Cyril SCETBON wrote:


>Jacques Caron wrote:
>>At 11:44 06/06/2008, Cyril SCETBON wrote:
>>>Hi guys,
>>>
>>>When I try to create a lot of set slonik dies with segmentation 
>>>fault message. The only way to succeed is to split the file 
>>>containing the create set statement and to try them one by one or 
>>>two by two. When I use strace on slonik it displays :
>>>
>>>read(3, "#--\n# Slony-I organizes tables i"..., 8192) = 8192
>>>read(3, "d table (set id=5, origin=201, i"..., 8192) = 8192
>>>read(3, ");\nset add table (set id=9, orig"..., 8192) = 8192
>>>mmap(NULL, 18446744073709547520, PROT_READ|PROT_WRITE, 
>>>MAP_PRIVATE|MAP_ANONYMOUS, -1, 0) = -1 ENOMEM (Cannot allocate memory)
>>
>>FYI, the mmap size is the (unsigned 64 bit integer) equivalent of 
>>-4096, so it looks like "something" attempted a negative size 
>>malloc (that got rounded to -4096, a multiple of the page size). 
>>Not quite sure which piece of code could lead to that result...
>the type is an unsigned integer. So if the value is too big a cast 
>occurs that should not give a negative value, but a shorter value...
>>
>>Jacques.
>>
>>_______________________________________________
>>Slony1-general mailing list
>>Slony1-general at lists.slony.info
>>http://lists.slony.info/mailman/listinfo/slony1-general
>
>--
>Cyril SCETBON - Ingénieur bases de données
>AUSY pour France Télécom - OPF/PORTAILS/DOP/HEBEX
>
>Tél : +33 (0)4 97 12 87 60
>Jabber : cscetbon at jabber.org
>France Telecom - Orange
>790 Avenue du Docteur Maurice Donat Bâtiment Marco Polo C2 - Bureau 202
>06250 Mougins
>France
>
>***********************************
>Ce message et toutes les pieces jointes (ci-apres le 'message') sont
>confidentiels et etablis a l'intention exclusive de ses destinataires.
>Toute utilisation ou diffusion non autorisee est interdite.
>Tout message electronique est susceptible d'alteration. Le Groupe France
>Telecom decline toute responsabilite au titre de ce message s'il a ete
>altere, deforme ou falsifie.
>Si vous n'etes pas destinataire de ce message, merci de le detruire
>immediatement et d'avertir l'expediteur.
>***********************************
>This message and any attachments (the 'message') are confidential and
>intended solely for the addressees.
>Any unauthorised use or dissemination is prohibited.
>Messages are susceptible to alteration. France Telecom Group shall not be
>liable for the message if altered, changed or falsified.
>If you are not recipient of this message, please cancel it immediately and
>inform the sender.
>************************************