Zeoslib Portal

Double post,

please update. I added a a dbc test for that report see https://sourceforge.net/p/zeoslib/code-0/6957/. Yet i had no time to port the test to to the dataset level. However please test and report your findings.

Or...... what happens if you kill the connection from a separate connection...? I'll test that out later on. You got me. I never ever had time to make tests for connection loss. Honestly i did some tests while running the test-suite, set a break-point, kill the server process, and test what happens...

Hi ae,

joarrrr such things are hard to test. Yes you assumption it's related is correct. For the connection lost issue i introduced IImmediatelyReleasable interface. Please test the fix. Could you consider to build a test which could interrupt a connection and we finally can test the behavior?

ae, no idea what the problem should be according the "encodings". Using TBytes has widely been advertised. Reason has been NEXT_GEN. Inbetween (XE10.4) NEXT_GEN is dead. And the logic is obsolate for me. If you need nice encoding methods. Look to Core\ZEncoding.pas According this thread: t...

Hi ash71, 1. please reply here: https://zeoslib.sourceforge.io/viewtopic.php?f=50&t=125819 we are waiting for an test to resolve the ticket: https://sourceforge.net/p/zeoslib/tickets/438/ 2. Is this regression related to: https://sourceforge.net/p/zeoslib/tickets/446/ ?? could you provide a simp...

Hello hoedlmoser,
if it's a TVarBytesField descendant:
http://docwiki.embarcadero.com/Librarie ... BytesField

if it's a fixed size TBytesField, i would agree..
So what's the field definition?

For the record. Since 7.3 i changed the memory consume of Zeos: Types having a fixed size are in a rowbuffer, doesn't matter if null or not. Changing this would kill the memory manager (to many small allocs). Variable types are allocated with Size: SizeOf(Pointer) in rowbuffer + Cardinal(Length)+Dat...

Hi billarchive, as Jan suggest, update from (SVN)-7.2-fixes branch first if you want stay in 7.2. An exception should be thrown, but a rounding of the fractions seems to be required. Note 7.3 (SVN)\trunk and \testing-7.3 already do support those exact time/Timestamp values including the invalid '000...

No i don't think i got you wrong. I'm just a step further to find root of evel. I'm trying to evaluate why the hell the data in the P := FResultSet.GetPWideChar(ColumnIndex, bLen); can be longer than the given field-Size. You can not comment the lines you suggest, or change the "truncation"...

Howdy Hardy, ahh that's clearing the things up. So it's no Zeos bug. Gerhard, FreeTDS is raw string based wheras OleDB is a UTF16 driver (SQLServer since 2005 is W based). Did you check what the SQL-Serverstudio shows for the values if you use FreeTDS? However OleDB is a W driver. Thus all encodings...

Hi i have commited a possible fix. (hope my crystal ball did catch the root of eval)
See: https://sourceforge.net/p/zeoslib/code-0/6849/

So please update and test. If the issue continues, please add the stacktrace.

Hello Gerhard, nope i have no clue yet. Seems like an encoding issue on either Driver or Zeos side(I'm forcing the driver for a full UTF16 conversion for all VARCHAR columns iirc).. We're using SQL-Server 2016 on a linux system using OleDB DelphiXE6 for production yet. JFYI Let's find the bug: 1. Wh...

marsupilami wrote: ↑23.09.2020, 12:40

EgonHugeist wrote: ↑23.09.2020, 06:20 No clue if i'll merge it to 7.2-fixes.

If it is simple, I vote for merging. Otherwise it has to go to the known issues list.

Done as requestet .. time for final 7.2 release, Jan

Hi hoedlmoser,

neither 7.1 nor 7.2 do support varbinary fields. 7.3 supports it. So my advice is: use 7.3.
@Jan we have no generic test for that. Could you add one to the suites?

Hi aehimself, feel free to make a patch for the module-name. A IZConnection.GetModuleName would be good i think. (What's the name for ADO/OleDB? we have no lib there..) According the FreeTDS-lib: Jan also maintains a compiled version in our SVN, check the \lib\FreeTDS folder.