7.2 testers-thread

[EgonHugeist]

@Miab3

got the little bugga. Ther was a wrong converions missing for String to Integer values in FireBird. R2840

But i still can not reproduce the issue the user is talking about...

[miab3]

@EgonHugeist,

R2841
It is still wrong(Firebird)

Code: Select all

 ZQuery1.Close;
 ZQuery1.SQL.Clear;
 ZQuery1.SQL.Add('SELECT count(*) as ile from zapisy where NR_KONTRA = :u');
 ZQuery1.ParamByName('u').AsString:='20';
 ZQuery1.Open;
 i:=ZQuery1.Fields[0].AsInteger; // <-- 0
 ZQuery1.Close;
 ZQuery1.ParamByName('u').AsInteger:=20;
 ZQuery1.Open;
 j:=ZQuery1.Fields[0].AsInteger; // <-- 17

In 7.1.2 helped in 7.2 it was not enough.

Michal

[EgonHugeist]

Michal,

sure? My test case:

Code: Select all

procedure ZTestCompCoreBugReport.TestParamUx;
var
  Query: TZQuery;
begin
  if SkipForReason(srClosedBug) then Exit;

  Query := CreateQuery;
  try
    // Query.RequestLive := True;
    Query.CachedUpdates := False;

    { Remove previously created record }
    Query.SQL.Text := 'DELETE FROM equipment WHERE eq_id>=:id';
    Query.ParamByName('id').AsInteger := TEST_ROW_ID - 2;
    Query.ExecSQL;

    Query.SQL.Text := 'INSERT INTO equipment (eq_id, eq_name) VALUES (:u, :u1)';
    Query.ParamByName('u').AsString := IntToStr(TEST_ROW_ID - 2);
    Query.ParamByName('u1').AsString := 'ab''cd''ef';
    Query.ExecSQL;
    Query.ParamByName('u').AsInteger := TEST_ROW_ID - 1;
    Query.ParamByName('u1').AsString := 'ab\cd\ef';
    Query.ExecSQL;
    Query.ParamByName('u').AsInteger := TEST_ROW_ID;
    Query.ParamByName('u1').AsString := 'ab\''cd\''ef';
    Query.ExecSQL;

    { Opens a result set. }
    Query.SQL.Text := 'SELECT * FROM equipment WHERE eq_id = :u or eq_id = :u +1 or eq_id = :u +2  ORDER BY eq_id';
    Query.ParamByName('u').AsString := IntToStr(TEST_ROW_ID-2);
    Query.Open;
    CheckEquals(TEST_ROW_ID - 2, Query.FieldByName('eq_id').AsInteger);
    CheckEquals('ab''cd''ef', Query.FieldByName('eq_name').AsString);
    Query.Next;
    CheckEquals(TEST_ROW_ID - 1, Query.FieldByName('eq_id').AsInteger);
    CheckEquals('ab\cd\ef', Query.FieldByName('eq_name').AsString);
    Query.Next;
    CheckEquals(TEST_ROW_ID, Query.FieldByName('eq_id').AsInteger);
    CheckEquals('ab\''cd\''ef', Query.FieldByName('eq_name').AsString);

    { Remove newly created record }
    Query.SQL.Text := 'DELETE FROM equipment WHERE eq_id>=:id';
    Query.ParamByName('id').AsInteger := TEST_ROW_ID - 2;
    Query.ExecSQL;
  finally
    Query.Free;
  end;
end;

It works perfectly for FB too.

[miab3]

@EgonHugeist, @mdaems,

Zeos7.2-alpha branches_ testing_R2844:
http://svn.code.sf.net/p/zeoslib/code-0 ... sting-7.2/
compiles and runs on (I have tested):

- Delphi XE2 32/64 -> C++(32)
Windows 8.1A pro-64 - C2D

I do not understand but for me also integer error in FB disappeared.
It would be worth this fix put in:
http://svn.code.sf.net/p/zeoslib/code-0 ... 1-patches/
(And Delphi XE5 package widh pathches)

Michal

[miab3]

@EgonHugeist, @mdaems,

Zeos7.2-alpha branches_ testing_R2849:
http://svn.code.sf.net/p/zeoslib/code-0 ... sting-7.2/
compiles and runs on (I have tested):

- Delphi XE2 32/64 -> C++(32)
Windows 8.1A pro-64 - C2D

Michal

[miab3]

@EgonHugeist,

R2853
Such a mistake is getting the second update (the first is successful).
Table Firebird.
In 7.0.6 with last ineger fix there is no error.

secupdfb.png

Michal

[EgonHugeist]

@miab3,

have my problems to understand you now. Do we still talk about the wrong FB string to int conversion?

If so than the SVN sequence is this:
1. development start in the testing branches
2. patches will be moved from testing to trunk (currently we didn't overwite the trunk, happens soon) in a weekly periode.
3. move 7.x signed patches to the X-patches branches too, right after applying the patches to trunk.

So we're sync with all branches and a new stable candidate can be assambled latet.

Or do i missunderstand you?

BTW: attached are two logs of my performance tests to compare the current work. The 7.2 ones is with a non commited Raw-logging patch. If we switch of the permanent logs it will change again a loads, i'm sure!

[miab3]

@EgonHugeist,

Do we still talk about the wrong FB string to int conversion?

No.
We are talking now about a new error in 7.2-alpha R2853,
which is clearly visible in the picture.
Additionally speech that this error does not occur in 7.1.2-stable with FB integer fix.

Michal

[EgonHugeist]

@miab3

thanx for the hint. Updating more than once did double all params... Fixed in 2854.
But that issue did going on a little bit longer.

Check my logs of previous post. The insertiation average did decrease a bit after applying the log-values patch.

[miab3]

@EgonHugeist, @mdaems,

Check my logs of previous post.

Nicely.

I noticed one thing.
Particularly slow are wide datasets (datasets with a large number of fields).
Slower than similar libraries.
Narrow are similarly fast.
It would be good to explain.

Update error disappeared

Zeos7.2-alpha branches_ testing_R2854:
http://svn.code.sf.net/p/zeoslib/code-0 ... sting-7.2/
compiles and runs on (I have tested):

- Delphi XE2 32/64 -> C++(32)
Windows 8.1A pro-64 - C2D

Michal

[EgonHugeist]

Particularly slow are wide datasets (datasets with a large number of fields).
Slower than similar libraries.
Narrow are similarly fast.
It would be good to explain.

Hard to explain, Michal. Didn't compare the speed with other components. It might be a architecture issue? Other libs mostly create direct access of the TCustomDataSets. Zeos has the dbc-layer too. Accessing data by using the dbc-layer only is lightning fast! Which doesn't mean we reach top of data yet. What i know: We're loads faster than UniDAC inbetween.

Right now I can't say if i can do some more accordingly fetching speed. It might be possible, have to think about it. (sometimes i have to go away from one task to refresh my mind and think about new ideas) For me the current refactoring is a huge success. Hope all IDE's support the newly introduced code.

If the logging refactoring is complete and function ZPlainString is out of use i'll write Arnaud Bouchez of Synopse again. I promissed him this rework on 7.1 but i couldn't do it because of missing time.

[miab3]

@EgonHugeist, @mdaems,

Zeos7.2-alpha branches_ testing_R2860:
http://svn.code.sf.net/p/zeoslib/code-0 ... sting-7.2/
compiles and runs on (I have tested):
- Delphi XE2 32/64 -> C++(32)
Windows 8.1A pro-64 - C2D

Michal

[miab3]

@EgonHugeist, @mdaems,

Zeos7.2-alpha branches_ testing_R2865:
http://svn.code.sf.net/p/zeoslib/code-0 ... sting-7.2/
compiles and runs on (I have tested):
- Delphi XE2 32/64 -> C++(32)
Windows 8.1A pro-64 - C2D

How should I fix MySQL timeout in 7.1.2?

Michal

[EgonHugeist]

@Miab3

a good question, since i'vo no idea if it realy works. And there is also a version issue for other Server-Arguments:
http://dev.mysql.com/doc/refman/5.0/en/ ... tions.html

until 5.0.12 the always used a PAnsiChar but this changed to the mysql generic types PByte, PLong etc.

Need to review this.

Michal, i've some settings issues for my performance tests. I always get 'MySQL has gone away' if i execute 'Delete from High_load' or 'Select * from High_load' and a huge number of rows. I could find a huge list of threads which all do point to a timeout issue. I'm not a MySQL hero so do you have an idea how to increase the timeouts for such actions? Or must i do something else?

[miab3]

@EgonHugeist, @mdaems,

Zeos7.2-alpha branches_ testing_R2877:
http://svn.code.sf.net/p/zeoslib/code-0 ... sting-7.2/
compiles and runs on (I have tested):
- Delphi XE2 32/64 -> C++(32)
Windows 8.1A pro-64 - C2D

Michael,
It is not a simple mistake?
http://sourceforge.net/p/zeoslib/tickets/52/

Michal