Errors, bugs, questions - page 1023

[Ilyas]

Zeleniy:

2013.07.30 07:27:55 Terminal OpenCL initialization skipped due to critical error, please update OpenCL drivers

OpenCL - what kind of drivers are these?

windows 8

One of OpenCL devices has driver error and terminal "crashes" when trying to use it.

Refer to [HKEY_LOCAL_MACHINE\SOFTWARE\Khronos\OpenCL\Vendors] registry branch to guess which OpneCL devices you have.

It is recommended to find updates for them.

[Mykola Demko]

shosh:

That's what I'm talking about!))

NormalizeDouble() -"Rounding a floating point number to the specified precision."

Well, if the mantissa remains the same size, what kind of rounding is it? It's just some kind of conversion, so to speak)))

Rounding is not even so much an abstract mathematical category, as a metrological one and implies exactly a reduction of the number of significant digits.

So this function is just a "imitator" of rounding, because the output is still the same double

It's not very convenient for me, because I needed it not for display but for calculations, but now I know how to deal with it.

Thank you very much for the clarification.

Wrong, there is a difference between rounding and outputting a value.

Output is affected by the output format, DoubleToString does the right format, rounding affects the mantissa digit values.

   Print(M_PI,"       //  PI без округления");
   Print(NormalizeDouble(M_PI,6),"       //  PI с округлением до 6-го знака");

  3.141592653589793       //  PI без округления
  3.141593                //  PI с округлением до 6-го знака

But the mantissa length remains the same since the variable type has not changed, it is still double.

[Renat Fatkhullin]

Zeleniy:

2013.07.30 07:27:55 Terminal OpenCL initialization skipped due to critical error, please update OpenCL drivers

OpenCL - what kind of drivers are these?

windows 8

There is a good article on installing OpenCL:How to install and use OpenCL in calculations

[Andrew Petras]

GPU Caps Viewer v.1.18.0 - free utility for OpenGL graphics cards

[Georgiy Merts]

Question: can't the terminal give something like an error code when an EX5 file fails to start?

Been fighting all day today - haven't beaten it, just "bypassed" the problem.

The project uses quite a complex class structure with a lot of objects, the system has grown quite large, but in several test EAs everything works fine.

Today I wanted to write a fairly simple Expert Advisor program, using old classes, it all compiled fine, but on startup it started writing "EX5 file error". I started gradually commenting all code fragments to find the error - no help. There were already empty functions in Expert Advisor - and it still wouldn't start. Deleted included mqh file - everything runs. I started to return the commented code - everything was ok, until I needed to use classes, declared in that include file... So, I understood, that just including mqh-file leads to errors EX5 (though compilation process goes without errors and warnings).

The only way to find out which part of the code causes such a strange error is to gradually comment out parts of the code, but this is easy only when the libraries are small. When libraries are large, one by one, this way is too cumbersome and dangerous (there is a probability that after finding an error not all the places will be uncommented again, which will cause further errors).

It would be good to understand at least approximately why the EX5 file does not start, although the compilation went without errors.

[Vasiliy Smirnov]

Laryx:

Question: Can't the terminal display some kind of error code when an EX5 file fails to start?

...

It would be nice to know at least approximately why the EX5 file didn't start, although it compiled without errors.

There's not much of a proof base.

[Georgiy Merts]

zfs:
Доказательной базы маловато.

I don't quite understand - what is there to prove? I was just expressing a wish, I think this point is important.

Well, if the developers think otherwise - well, whatever, let's do without it...

[Vasiliy Smirnov]

Laryx:

I don't quite understand - what is there to prove? I was just expressing a wish, I think this point is important.

Well, if the developers think otherwise - well, whatever, let's do without it...

You need logs, libraries, exefiles and to servicedesk.

[Roffild]

Roffild:
A simple indicator:

I throw it on the M5 chart:

2013.07.26 22:17:12 testprevbar (EURUSD,M5) Full recalculation!
2013.07.26 22:17:12 testprevbar (EURUSD,M5) Full recalculation!
2013.07.26 22:17:06 testprevbar (EURUSD,M5) Full recalculation!
2013.07.26 22:17:00 testprevbar (EURUSD, M5) Full recalculation!

https://www.mql5.com/ru/forum/1111/page1038#comment_554925

This bug is reproduced when throwing the indicator on a chart opened from a tester (2010 for example).

I have already described the actions in the bug report.

Until this bug is fixed, do not use the open chart from the tester in real trading.

[Vasiliy Smirnov]

Roffild:

And until this bug is fixed, do not use the open chart from the tester in real trading.

Is this a bug in your program? Is it allowed to use tester charts for real trading?