7.2 testers-thread

[miab3]

@EgonHugeist, @mdaems,

Zeos7.2-alpha branches_ testing_R2914:
http://svn.code.sf.net/p/zeoslib/code-0 ... sting-7.2/
compiles and runs on (I have tested):

- D2006,
- DXE2 32/64 -> C++(32),

Firebird 2.5.3, MySQL 5.5.18, MariaDB 5.5.32, PostgreSQL 9.3.0.1, MSSQL 2008 R2 Express(mssql, FreeTDS, ado-sqlncli), *.mdb-ado-Jet.OLEDB.4.0, SQLite 3.7.17.

Referring to Ticket54.
With TESTDB.FDB.

Code: Select all

procedure TForm1.Button1Click(Sender: TObject);
begin
  Query.Close;
  Query.SQL.Text := 'insert into tbl(num) values (8) returning id';
  Query.Open;
end;

This works.

But Button2Click after Button1Click.

Code: Select all

procedure TForm1.Button2Click(Sender: TObject);
begin
  Query.Close;
  Query.Open;
end;

causes an error:

fb_ret.png

Michal

[EgonHugeist]

@Miab3

made a reply: https://sourceforge.net/p/zeoslib/tickets/54/

[miab3]

@EgonHugeist,

Michael,
Why in Zeos 7.1.2 and Zeos 7.2. x
for Firebird table:

Code: Select all

CREATE TABLE TEST3 (
  F1 DECIMAL(10, 0),
  F2 VARCHAR(20),
  F3 BLOB);

For field F1:
- On close dataset 'Add Fields...' add TLargeintField type (without Error),
- On open dataset 'Add Field...' add TFloatField with Error:

FB_DECIMAL10.png

Michal

[miab3]

@EgonHugeist,

Zeos 7.2.x_R2914
There is something wrong with the:
{$Define WITH_ZSTRINGFIELDS}
Sometimes the value of a field in the grid is not refreshed only reproduced the first.

For example, such SQL(Firebird):

Code: Select all

Select RDB$RELATION_NAME from RDB$RELATIONS WHERE (RDB$VIEW_BLR is NULL) AND (RDB$SYSTEM_FLAG = 0) ORDER BY 1

Michal

[EgonHugeist]

@Miab3

both issues confirmed. I'll check them now. The second with the TZFields is a bit wired. If you click to another row an move the mouse coursor ... Than always the current value of the row came to shove. Have to check if a row index is available too..

[Zoran]

@Zoran

lorbs, what's that? It compiles well on my FPC2.6.2 LCL see attachment. Which environmet do you use? Did you compile ZCore.lpk first?

Actually, StrLen is defined in System, as well as in SysUtils.

Still, package ZCore cannot compile, the compilation breaks on this line with message:

.../zeos/testing-7.2/src/core/ZFastCode.pas(5804,64) Error: Incompatible types: got "strlen(PChar):Int64;" expected "<procedure variable type of function(PChar):LongInt;Register>"

This is on Linux Mint 15, 64 bit, but later I'm going to try on Windows 7 too and I will post here.

[Zoran]

This is on Linux Mint 15, 64 bit, but later I'm going to try on Windows 7 too and I will post here.

Same on Windows, but only with 64-bit compiling!

C:\lazarus\Komponente\Zeos\testing-7.2\src\core\ZFastCode.pas(5804,64) Error: Incompatible types: got "strlen(PChar):Int64;" expected "<procedure variable type of function(PChar):LongInt;Register>"

When compiling in 32-bit, everything goes fine.

[EgonHugeist]

@Zoran

thanks for your hints. Did add a define. Can you please test it?

@Miab3

hope i got the TZStringField issues resolved. Some optimations can be done. Didn't had the time to check the TLargeInt vs. TNumeric-Field thing yet...

[Zoran]

@Zoran

thanks for your hints. Did add a define. Can you please test it?

Now, zcore compilation breaks with

../zeos/testing-7.2/src/core/ZEncoding.pas(1265,1) Fatal: Syntax error, "identifier" expected but "BEGIN" found

Ok, this one was easy, the "var" keyword from line 1260 should be moved down, inside the directive. So, I did that and tried to compile. Now zcore compiles. Then, zplain, zparsesql also compiles. But zdbc does not:

../zeos/testing-7.2/src/dbc/ZDbcMySqlResultSet.pas(416,48) Error: Call by var for arg no. 2 has to match exactly: Got "LongWord" expected "QWord"

and later in same unit:

../zeos/testing-7.2/src/dbc/ZDbcMySqlResultSet.pas(781,39) Error: Call by var for arg no. 2 has to match exactly: Got "LongWord" expected "QWord"

[EgonHugeist]

@Zoran

again thank you. Did download a 64Bit FPC now and i think i did resolve the remaining issues. (Wondering why FPC assume a UInt64/QWord for the type LongWord)

Patch done R2922. Please test it again.

@Miab3
found another silly stup with the new Fields. I couldn't update them. Resolved too. I'll go forward and write a TZ()Field descendant for each type (our conversions are faster then these of the Delphi/FPC rtl and check the benchmarks again). The Synopse benchmarks... Sade yes he really didn't use 7.2 and he also needs to rework his framework for 7.2 in some case. I'll support him a bit more if Mark is ready with the Logging patches.

[Zoran]

@Zoran
again thank you. Did download a 64Bit FPC now and i think i did resolve the remaining issues.

Now, compiles well on Windows 64-bit. I'll try in Linux this evening.

@Zoran
(Wondering why FPC assume a UInt64/QWord for the type LongWord)

What do you mean? This:

Code: Select all

procedure TForm1.Button1Click(Sender: TObject);
begin
  Edit1.Text := 'Size of LongWord=' + IntToStr(SizeOf(LongWord))
                 + '  Size of UInt64=' + IntToStr(SizeOf(UInt64))
                 + '  Size of QWord=' + IntToStr(SizeOf(QWord));
end;

returns

Size of LongWord=4 Size of UInt64=8 Size of QWord=8

When compiled with either 32-bit or 64-bit FPC (just as documented).

Anyway, instead of using compiler directives, you should use PtrUInt type (pointer sized, unsigned, integer type) where needed.
This:

Code: Select all

procedure TForm1.Button1Click(Sender: TObject);
begin
  Edit1.Text := 'Size of PtrInt=' + IntToStr(SizeOf(PtrInt))
                 + '  Size of PtrUInt=' + IntToStr(SizeOf(PtrUint));
end;

when compiled with 32-bit compiler, returns

Size of PtrInt=4 Size of PtrUInt=4

but when compiled with 64-bit compiler, it returns

Size of PtrInt=8 Size of PtrUInt=8

[moctes]

Hi to all,

I ran Arnaud Bouchez mORMot benchmarks, see http://synopse.info/forum/viewtopic.php?id=1466 and http://synopse.info/forum/viewtopic.php?id=1467 using latest Zeos 7.1 stable, and I'm curious about the 7.2 branch, Do you think that if I do the test with the latest version will be a noticeable improvement? and before doing anything, Would you say 7.2 alfa is stable enough for this kind of tests?

Regards and thank you for your work.

[EgonHugeist]

@Zoran,

thanks for you hints. I already know that. I was just curious about the PUlong trouble-maker. Thanks for compiling confirmation. Hope you check against performance too?

@moctes

I'm curious about the 7.2 branch

Why? 7.2 is a pure performance rework because of the Synopse tests..

Do you think that if I do the test with the latest version will be a noticeable improvement?

Absolutely! I did increase fetching and insertiation speed in a unexpected high rate for the most protocols. Some remaining things need to be done.

Would you say 7.2 alfa is stable enough for this kind of tests?

Sure. 7.2 had some hickups which (hopefully all) are resolved yet. But making the Synopse test with AB's current access won't bring some other results i think. He need some framework reworks for the ResultSets and the Statements. We'have no final state yet. But it would be a good idea to start reowrking the Synopse<>ZDbc interface. Also are there some Prameters for pimping the speed.
As you can read in this thread: I planed to write him and i'm willing to support him. Mayby you give him a hint? I promissed the speed upgrades to 7.1 but i had other things todo. So we decided to release 7.1 and make it on 7.2. Actually MArk is preparing a patch to prevent value logging (a remaining perforance Killer) and i do Rework the TFields because the TDataSet interface could work more powerfull if we have differet access here..
Also do i plan to add some more TVariant (mainly pointing val's like PInteger..) types and some more exact TZSQLTypes (stSmall is already done now).
The IZLobs got also a rework. Uncached-Lobs (load on demand) are added where i think it was possible. Also do i differ between BLob and Clob.

But trust me the actual state really does perform loads (mostly x3 for the UnicodeIDE's f.e.)
.... There is mutch to say about. Can you check the Synopse-Framwork and make the neccessary patches for A.Bouchez? Than i'm willing to support you where ever i can..

[moctes]

@EgonHugeist,

Great!! I'm preparing to make a 2 day work travel, but I'll see if I can test it today, otherwise it will be until friday and I will keep you both informed of my findings (AB and You)

Best regards.

Mocte

[miab3]

@EgonHugeist,

Michael,
But TBytes appears from Delphi 2007.
type TBytes = array of Byte;

Michal