Errors, bugs, questions - page 1625

 
Vladislav Andruschenko:

mt4 compiles in 888 msec.

the same project compiles 4103 ms in mt5

Well, it seems that the problems are connected with v5. 4.5 difference is not a little. Although to complete the picture it would be nice to check with old compiler (1159), if you have it (or I can download it). But maybe you will have to make some corrections in your code.
 
-Aleks-:

I don't have money for MT5 - I trade cent accounts and the DC is in no hurry to open them for a fiver.

Not looking hard enough. https://www.mql5.com/ru/forum/88768/page2#comment_2587760
Крупнейшие брокеры отмечают взрывной рост популярности MetaTrader 5
Крупнейшие брокеры отмечают взрывной рост популярности MetaTrader 5
  • reviews: 3
  • www.mql5.com
Недавно один из национальных брокеров России Solid Financial Services запустил торговую платформу MetaTrader 5 с хеджинговой системой учета позиций...
 
Alexey Navoykov:

I would like to raise the issue of slow MQL5 compilation once again. About three months ago I tried to raise this issue, but it was not understood, apparently my arguments were not convincing enough. Therefore, I rolled back to the old build (1159), which compiled everything almost instantly (while with new compilers my project compiled in 20 seconds).

And so a week ago I've been trying to switch to a new build. I figured "forget about 20 seconds, I'll bear with it for the sake of new stuff". Of course I had to tweak the code a bit to comply with new conditions, which revealed several bugs of the new compiler (described here).The upshot is that my project has been compiling for 30 seconds already! I don't know if it has to do with the complication of the project or with one more "complication" of the compiler, but it just doesn't fit anymore.

The project contains about 700 Kb of source code, it's an Expert Advisor containing a couple of dozens of mqh. Everything is OOP. People wrote to me earlier that the slowdown is probably caused by large functions. I've had a few of them. Well, I've fragmented them into smaller parts and they have no effect.

What's the most astonishing is that this superlong compilation is of no use at all. The speed of the program is the same as with the old compiler, I measured it specifically. That begs only one phrase: "What for?".

I have a strong feeling that there is a bug/malfunction in the compiler because of which it is idly racing through an empty space. How else can I explain the fact that an absolutely empty script with only the OpenStart() { } function compiles over 400 ms!It is unimaginable that it can take so long to compile/optimize an empty script. Well, by adding small functions and classes into it, you can see how rapidly the compilation time grows.

I want to say right away that my hardware is of course far from powerful - Core i5U. But this doesn't prevent my project from compiling in 1-2 seconds on an old compiler. Respectively, the dummy is compiled there in a moment.

I'll also note. The compiler completely lacks not only caching of fragments compiled earlier but even a trivial check to make sure that the source code was identical. I.e. you compile your project and then click the "Compile" button again without making any changes and again wait for the same 30 seconds. How good is it...

I would like to hear comments from MT developers and from forum users, working with large projects (is it just me who has this problem?), how long it takes to compile, etc. We must say right away that we are talking about compiling an executable.

My projects have more than a dozen of source files, like yours, and all on OOP, while I will not claim about 20 seconds, but more than 11 - 14 seconds I see constantly. Yet some kind of caching is taking place, since if you don't change anything the time is changing by 1-2 seconds in any unpredictable direction. I'm not comparing the build of the projects using the old and new compilers because the old compilers built everything much faster. I think the developers themselves see this point and someday they will fix it :) It's not for nothing that they release several new runs every month - it means they see something and fix it.

 

Terminal version and bit rate

v.1375, 64-bit

Problem description.

After upgrade to the latest build, agents freeze after passing the first 1900-2100 passes during optimization. Everything was fine before update, all parameters and EA code are the same.

Sequence of actions

Optimization starts. Opening broker. Real account. The tools: Si Splice, Vtb Splice, Si 9.16, Vtb 9.16 (I have not tried others). Interval: monthly, minute, 15-minute. Opening prices or OHLC.

Result.

Local and remote agents after 2000 passes actually freeze, CPU loads, change by about 0.01% per 10 minutes. 14 agents.

Expected result

Optimization walkthrough as with the previous build.

Additional information

About me: experienced .net MQL5 programmer


I have looked at the logs everywhere. I have compared them to the logs of the previous build. I have not encountered any problems or errors. Quality of the history is good.
Files:
image.jpg  89 kb
 
coderex:

However, there is some caching going on, because if you don't change anything, the time changes by 1-2 seconds in any unpredictable direction.

:) it's obviously just an error, influence of other processes in the system, plus disk cache. In any case, 10-15% is not at all the indicators, for which caching is done.

I wouldn't compare building projects with old and new compilers because old compilers built everything much faster. I think developers see this point and will fix it at some point :) It's not for nothing that several new runs are released every month, it means they see something and fix it.

New betas are released because people complain about various bugs, but if bugs are a solid argument for them to fix, all the rest ... You have to convince them for a long time. Even when it would seem clearly brought all the arguments, clearly described the picture as clear as can be, they still resist, by hook or by crook :) Here I am the same 3 months ago tried to convince them, but no one supported.

And I've seen that only a few people use MQL for big projects and they probably don't bother with small ones because of an extra couple of seconds.

By the way, what kind of CPU do you have?

 
Alexey Navoykov:

... they're going to do whatever they can :)

That's why I'm not trying to prove anything :) In addition, pluses projects take much longer to build, although they are much larger, but I'm used to pluses builds in minutes for executable or library file, while the project of several files with directory structure takes up to several tens of minutes :) and to wait 10-20 seconds is not a problem...
 
Alexey Viktorov:
Didn't look hard enough. https://www.mql5.com/ru/forum/88768/page2#comment_2587760
The link does not give the information of interest - be specific.
 
ProfitTraderRU:

Terminal version and bit rate

v.1375, 64-bit

Problem description.

After upgrade to the latest build, agents freeze after passing the first 1900-2100 passes during optimization. Everything was fine before update, all parameters and EA code are the same.

Sequence of actions

Optimization starts. Opening broker. Real account. The tools: Si Splice, Vtb Splice, Si 9.16, Vtb 9.16 (I have not tried others). Interval: monthly, minute, 15-minute. Opening prices or OHLC.

Result.

Local and remote agents after 2000 passes actually freeze, CPU loads, change by about 0.01% per 10 minutes. 14 agents.

Expected result

Optimization walkthrough as with the previous build.

Additional information

About me: experienced .net MQL5 programmer


I have looked at the logs everywhere. I have compared them to the logs of the previous build. I have not encountered any problems or errors. The quality of history is good.

Does this behavior occur with any Expert Advisor?

It would be nice to have the logs. Please submit a ticket to Service Desk.

 
I only tested my EAs. With previous build they were optimised normally.

I have madean application to servicedesk. Believe me, there is nothing unusual in the logs (saw logs of previous and current build).
 
Alexey Da:

Terminal logs are good, more interesting are logs of strategy tester and testing agents.

+ attach to your ticket at least EX5 of your Expert Advisor (we will delete it after research) + description of parameters used during optimisation.

OK, thank you.
Reason: