Thread: invalid byte sequence
After upgrading to pg 9.0.3 (from 8.4.2) on my Mac OS 10.6.2 machine i find this in my log file (a lot): <postgres%192.168.254.210%2011-03-03 16:37:30 CET%22021>STATEMENT: SELECT pg_file_read('pg_log/postgresql-2011-03-03_000000.log',250000, $ <postgres%192.168.254.210%2011-03-03 16:37:32 CET%22021>ERROR: invalid byte sequence for encoding "UTF8": 0xe3bc74 Apparently pg doesn't like the contents of that logfile. The folks from the pgadmin list (i noticed the problem using the pgadmin logviewer and asked for help there) advised me tochange LC_messages locale to 'C' (i had it on 'de_DE-UTF8' before), but that doesn't appear to help. The Server encodingis UTF8. No special client encoding is set. Any help would be appreciated, Max Maximilian Tyrtania Software-Entwicklung Dessauer Str. 6-7 10969 Berlin http://www.contactking.de
On 04/03/11 00:02, Maximilian Tyrtania wrote: > After upgrading to pg 9.0.3 (from 8.4.2) on my Mac OS 10.6.2 machine i find this in my log file (a lot): > > <postgres%192.168.254.210%2011-03-03 16:37:30 CET%22021>STATEMENT: SELECT pg_file_read('pg_log/postgresql-2011-03-03_000000.log',250000, $ > <postgres%192.168.254.210%2011-03-03 16:37:32 CET%22021>ERROR: invalid byte sequence for encoding "UTF8": 0xe3bc74 The "0xe3bc74" looks like gibberish in any encoding I can think of. What's the input file? Is it sanely encoded? Do you know what encoding it is in? If you really want to be encoding-agnostic and you do not care if you get garbage data in your database that makes no sense and can never make any sense, then you must ensure that your database is in the "C" locale for LC_CTYPE and LC_COLLATE, and you must SET client_encoding = "SQL_ASCII" when reading the data. A suitable CREATE DATABASE command might be: CREATE DATABASE garbage TEMPLATE template0 ENCODING 'SQL_ASCII' LC_COLLATE 'C' LC_CTYPE 'C'; but I really don't think that's generally a good idea. Storing random crap in text fields will cause you pain later. Better to either convert the text to a sane encoding, store it as bytea if you want the raw bytes, or reject it. -- Craig Ringer
Am 04.03.2011 um 11:01 schrieb Craig Ringer: > On 04/03/11 00:02, Maximilian Tyrtania wrote: >> After upgrading to pg 9.0.3 (from 8.4.2) on my Mac OS 10.6.2 machine i find this in my log file (a lot): >> >> <postgres%192.168.254.210%2011-03-03 16:37:30 CET%22021>STATEMENT: SELECT pg_file_read('pg_log/postgresql-2011-03-03_000000.log',250000, $ >> <postgres%192.168.254.210%2011-03-03 16:37:32 CET%22021>ERROR: invalid byte sequence for encoding "UTF8": 0xe3bc74 > > The "0xe3bc74" looks like gibberish in any encoding I can think of. > What's the input file? We are talking about pg's own logfile here. I thought that was clear. Look at the file's name. Apparently some guy on thefrench pgAdmin list has the very same problem. I have no idea how "0xe3bc74" made it into the log file. > Is it sanely encoded? Do you know what encoding > it is in? As i said, i initially set lc_messages to 'de_DE-UTF8', so i assume that's what the log file was in. I changed it to 'c'now. > If you really want to be encoding-agnostic and you do not care if you > get garbage data in your database that makes no sense and can never make > any sense, then you must ensure that your database is in the "C" locale > for LC_CTYPE and LC_COLLATE, and you must SET client_encoding = > "SQL_ASCII" when reading the data. > > A suitable CREATE DATABASE command might be: > > CREATE DATABASE garbage > TEMPLATE template0 > ENCODING 'SQL_ASCII' LC_COLLATE 'C' LC_CTYPE 'C'; > > but I really don't think that's generally a good idea. Storing random > crap in text fields will cause you pain later. Better to either convert > the text to a sane encoding, store it as bytea if you want the raw > bytes, or reject it. I certainly don't want to be encoding agnostic. I just would like to be able to read my log file using PGAdmin, which i can'tright now, because PGAdmin 1.12. chops off the content after the 1st character that doesn't match the encoding. Best wishes, Max Maximilian Tyrtania Software-Entwicklung Dessauer Str. 6-7 10969 Berlin http://www.contactking.de
On 4/03/2011 10:18 PM, Maximilian Tyrtania wrote: > Am 04.03.2011 um 11:01 schrieb Craig Ringer: > >> On 04/03/11 00:02, Maximilian Tyrtania wrote: >>> After upgrading to pg 9.0.3 (from 8.4.2) on my Mac OS 10.6.2 machine i find this in my log file (a lot): >>> >>> <postgres%192.168.254.210%2011-03-03 16:37:30 CET%22021>STATEMENT: SELECT pg_file_read('pg_log/postgresql-2011-03-03_000000.log',250000, $ >>> <postgres%192.168.254.210%2011-03-03 16:37:32 CET%22021>ERROR: invalid byte sequence for encoding "UTF8": 0xe3bc74 >> >> The "0xe3bc74" looks like gibberish in any encoding I can think of. >> What's the input file? > > We are talking about pg's own logfile here. I thought that was clear. Look at the file's name. Apparently some guy on thefrench pgAdmin list has the very same problem. I have no idea how "0xe3bc74" made it into the log file. Oh. Good point. PostgreSQL's logs suffer from an interesting bug in situations where the server uses mixed encodings, causing the log files to contain text with more than one encoding. There's been prior discussion of it, but no conclusions. > >> Is it sanely encoded? Do you know what encoding >> it is in? > > As i said, i initially set lc_messages to 'de_DE-UTF8', so i assume that's what the log file was in. I changed it to 'c'now. Well, the PostgreSQL log files are *not* sanely encoded, so that's part of the problem. -- Craig Ringer Tech-related writing at http://soapyfrogs.blogspot.com/