Thread: ...
Hi, just to report a strange behaviour in PostgreSQL v6.3.2, Linux 2.0.34. I do this: hjk=> alter table control_veh add column fecalta date; ADD then that: hjk=> update control_veh set fecalta='01-01-1800' where plact='440JAS'; ERROR: Relation control_veh does not have attribute fecalta later this: hjk=> \d control_veh Table = control_veh +----------------------------------+----------------------------------+-------+ | Field | Type | Length| +----------------------------------+----------------------------------+-------+ | paterno | char() | 25 | ...many fields... | ultpago | char() | 2 | | fecalta | date | 4 | +----------------------------------+----------------------------------+-------+ this again: hjk=> update control_veh set fecalta='01-01-1800' where plact='440JAS'; ERROR: Relation control_veh does not have attribute fecalta desperate this: hjk=> vacuum control_veh; VACUUM hjk=> alter table control_veh add column fecalta date; ERROR: PerformAddAttribute: attribute "fecalta" already exists in class "control_veh" hjk=> then again: hjk=> update control_veh set fecalta='01-01-1800' where plact='440JAS'; UPDATE 1 hjk=> select * from control_veh where plact='440JAS'; and voil�, it's complete. Isn't it strange? -- La Mancha, http://breogan.iimas.unam.mx/~mancha casa://AvRevolucion.1761.depto.5/~mancha ring://550-2547.df.telmex.com.mx/~pedir.por.mancha chamba://cubo-320.iimas.unam.cu/~mancha rechamba://533-3035.df.telmex.com.mx/~pedir.por.Sr.Magari�os
LMdlCqL> Hi, just to report a strange behaviour in PostgreSQL v6.3.2, Linux LMdlCqL> 2.0.34. LMdlCqL> I do this: hjk=>> alter table control_veh add column fecalta date; LMdlCqL> ADD LMdlCqL> then that: hjk=>> update control_veh set fecalta='01-01-1800' where plact='440JAS'; LMdlCqL> ERROR: Relation control_veh does not have attribute fecalta LMdlCqL> later this: hjk=>> \d control_veh LMdlCqL> Table = control_veh LMdlCqL> +----------------------------------+----------------------------------+-------+ LMdlCqL> | Field | Type | Length| LMdlCqL> +----------------------------------+----------------------------------+-------+ LMdlCqL> | paterno | char() | 25 | LMdlCqL> ...many fields... LMdlCqL> | ultpago | char() | 2 | LMdlCqL> | fecalta | date | 4 | LMdlCqL> +----------------------------------+----------------------------------+-------+ LMdlCqL> this again: hjk=>> update control_veh set fecalta='01-01-1800' where plact='440JAS'; LMdlCqL> ERROR: Relation control_veh does not have attribute fecalta LMdlCqL> desperate this: hjk=>> vacuum control_veh; LMdlCqL> VACUUM hjk=>> alter table control_veh add column fecalta date; LMdlCqL> ERROR: PerformAddAttribute: attribute "fecalta" already exists in class "control_veh" hjk=>> LMdlCqL> then again: hjk=>> update control_veh set fecalta='01-01-1800' where plact='440JAS'; LMdlCqL> UPDATE 1 hjk=>> select * from control_veh where plact='440JAS'; LMdlCqL> and voilà, it's complete. Isn't it strange? This is an pgsql's old bug already reported by me. The problem is that you can't reproduce this behavior all the time. I have this bug also, but it appears only sometimes. It isn't danger only disquieting. ;-) Buena suerte! Jose'
>This is an pgsql's old bug already reported by me. >The problem is that you can't reproduce this behavior all the time. >I have this bug also, but it appears only sometimes. It isn't danger >only disquieting. ;-) Yes, it doesn't interferes with my sleeping, but as that particular table is growing to near 6 million registers, each vacuum takes more that I'd like. Which brings another question: has any PC Linux user applied the patches so i386 Linux can use the 64 bits filesystem? This is because we now have tables than soon be over the 2 GB in size. -- La Mancha, http://breogan.iimas.unam.mx/~mancha casa://AvRevolucion.1761.depto.5/~mancha ring://550-2547.df.telmex.com.mx/~pedir.por.mancha chamba://cubo-320.iimas.unam.cu/~mancha rechamba://533-3035.df.telmex.com.mx/~pedir.por.Sr.Magari�os