[Monetdb-developers] overflow checks in the Pathfinder shredder?
Jan F. (and/or others),
did I understand you correctly on Tuesday that there are checks in the pathfinder shredder (XQuery_0-12 release branch) that detect overflows in case of huge documents, i.e.,
level > GDK_chr_max size > GDK_int_max OIDs > GDK_oid_max
and generate the respective errors?
If not, shouldn't we add them, and trigger a proper error in case they occur?
The same question arises also for (intermediate) results during query processing...
Stefan
Stefan,
We check the main PRE number against INT_MAX and we check if the the size of a BAT does not try grow beyond INT_MAX. We do not check the 'level' and 'size' values against the max values you mentioned. I will add these checks and change the max value for PRE numbers from INT_MAX into GDK_oid_max,
Jan.
On Wednesday 21 June 2006 13:39, Stefan Manegold wrote:
Jan F. (and/or others),
did I understand you correctly on Tuesday that there are checks in the pathfinder shredder (XQuery_0-12 release branch) that detect overflows in case of huge documents, i.e.,
level > GDK_chr_max size > GDK_int_max OIDs > GDK_oid_max
and generate the respective errors?
If not, shouldn't we add them, and trigger a proper error in case they occur?
The same question arises also for (intermediate) results during query processing...
Stefan
Jan,
sounds great. thanks!
btw:
watch out that tests like (x > GDK_int_max) only work properly if x has a larger domain than int! otherwise, either the compiler already complains, or the result should always be false...
to be on the save side, we should better use (x < GDK_int_max) or (x >= GDK_int_max).
likewise for the other types.
Stefan
On Wed, Jun 21, 2006 at 01:47:48PM +0200, Jan Flokstra wrote:
Stefan,
We check the main PRE number against INT_MAX and we check if the the size of a BAT does not try grow beyond INT_MAX. We do not check the 'level' and 'size' values against the max values you mentioned. I will add these checks and change the max value for PRE numbers from INT_MAX into GDK_oid_max,
Jan.
On Wednesday 21 June 2006 13:39, Stefan Manegold wrote:
Jan F. (and/or others),
did I understand you correctly on Tuesday that there are checks in the pathfinder shredder (XQuery_0-12 release branch) that detect overflows in case of huge documents, i.e.,
level > GDK_chr_max size > GDK_int_max OIDs > GDK_oid_max
and generate the respective errors?
If not, shouldn't we add them, and trigger a proper error in case they occur?
The same question arises also for (intermediate) results during query processing...
Stefan
participants (2)
-
Jan Flokstra -
Stefan Manegold