I am finding that after about 3988 HD commands, I stop receiving data. It happens every time. It does not seem to matter which symbols they are. If I take the exact same code, and simply replace it with a HM command, it all works fine. Also the HD commands seem much slower that HM commands.

I am running the latest 4.0 release.

---

sruff

Does it always barf on the same symbol? Or does the 3988'th request always hang regardless of the symbol?

I don't believe it is symbol dependent. I did the test starting at an offset of 2000 in my symbol list and it still hung up at about the same time, but I will trace which symbols it hangs on and see if removing them helps.

Also I need to correct an earlier statement that the HM command worked fine. It worked longer but eventually hung at about 5000 symbols. If I let it just sit there after it hangs, IQConnect eventually crashes.

---

sruff

I have 4.00.2 hanging on many successive HT requests. The code was working perfectly with 2.3.0.1 for many months, and now after many symbol requests the socket hangs (I am using java to register the client and then TCP/IP). In fact, when this occurs I cannot access the network with any program on the IQConnect computer (I receive a "no buffer space available" network error from other programs when trying to access the network. If I restart IQConnect (by stopping it in the taskbar causing my program to re-niitiate a connection) everything is fine until several more symbols and it hangs again.

Please make this a high priority as historical downloading does not appear to work in 4.0.2 for more than a few symbols.

Thank You,
Scott

It would be helpful if you could identify how it crashes. Are you using a multi-threading model? Their new version can wreak havoc if you don't have multi-threading implemented properly. I had to do a bug-fix on my code to make it work properly.

You might also check your error code handling. I know they've made some changes in that regard. Look for messages like:

!ERROR! !NONE!
!ERROR! Invalid symbol.

I've been using 4.0.0.2 in a C# multi-threaded situation for the last week or two on 2600 symbols or so with 100 simultaneous HD's, and I have had 100% responsiveness with no crashes.

In addition, if you are using their mktsymbol.txt file as a seed, last time I looked, about 10 days ago, it has one problem line in it: there is a wrapped line that they don't seem to be fixing. If your symbol import doesn't catch that line, it may wreak havoc on the remainder of your symbol import/handling code.

Thanks for the help. I'll take another look at my code. This could possibly be related to http://forums.iqfeed.net/index.cfm?page=topic&topicID=656 because I had to put code in to "clean out " the socket after an invalid symbol request after migrating from 2.3.0.1.

Will post if the lockup still continues to appear in IQConnect. Thanks

Hi,

Please let me know if you are still seeing the lockup, and I can look into it. Thanks!

Natalie

---

Natalie Hannan DTN Market Access, LLC.

Natalie,

I am still getting the hang. For me it happens even earlier if I use only one thread. When it hangs up I need to kill IQConnect to even browse the internet. It is not symbol related because I filled my entire symbol vector with "INTC", and the problem still occurs.

However, I have not completely ruled out a problem in my code. My hope is to create a very simple, single threaded console app that repros the problem. I will then test that on a couple of machines, and a wired and wireless network. If I can repro everywhere with the simple app, I will send it to you.

Thanks for your reply!
Scott

---

sruff

Hi Scott,

That sounds good. If you can reporduce the issue, please send the app to the developer support email. Thanks!

Natalie

---

Natalie Hannan DTN Market Access, LLC.

>>scruff wrote: When it hangs up I need to kill IQConnect to even browse the internet

I've reviewed my code and still receive this exact behavior when downloading a large number of symbols with HT requests (works fine for small number of symbols < 500 or so, even if some requests are for bad or empty symbols). It is a single-threaded java and TCP/IP implementation. As I mentioned this code has worked well with 2.3.0.01 for many months. The only changes have been to accomodate new messages/ syntax on 4.0 and the synchronization issues referenced on another thread in this forum.

The program hangs in different places each time rendering all network access from that machine impossible until IQConnect is killed. Since others are having this problem I do believe the problem is with IQConnect. If scruff is unable to provide the demo mentioned I'll see what I can do to provide one. Thanks

I have completed my simple test app. I have tested on two machines on two different networks, both wired and wireless, and I still get the hang. My test program is a simple single threaded console app that sends the HD command for INTC asking for three days of data. It always asks for INTC to eliminate any symbol issues. It dumps to standard out the count of the request and all data read from the socket. It also should dump any errors sending or reading from the socket, but I'm not seeing any errors.

I will send a zip file containing the project to the support email.

---

sruff

Thanks for putting together the test app sruff. As I posted I am also having this problem and look forward to a DTN resolution as soon as possible.

I believe I am seeing the same thing. I'm not sure what my symbol count is, but it may be in the approx 3800 symbol area.

I run about 2500 symbols to get eod data. Because of what I think are some drop outs, I'll rerun it again. The run stops 'half way through'. I'd have to do some checking to figure out exactly how many, but 2500 plus 'half' would be around 3800.

When this happens, I simply exit my program and restart it. I don't have hard freezes like others in this thread indicate (I'm using C# with tcp/ip socket connections). Therefore, the lock ups appear to be session based.

What are the results of the sample ap that was submitted by someone earlier in this thread? Has it lead to any conclusions yet?

Based upon comments so far, and some of the errors I've seen in logging, I'm guessing there are still some subtle bugs in the new v4 code.

For the drop outs I mention earlier, I get error messags like:

CPN !ERROR! !NONE!
CSE !ERROR! Invalid symbol.

Each run servers up random errors like this. As far as I can tell, never the same symbol, hence multiple rungs to get full coverage.

These errors could be due to my reusing buffers when I shouldn't be, or it could be the v4 code not processing buffers properly. I seem to recall that my code worked back in the early days of v2, so I'm thinking it is a v4 problem (until I'm proven wrong!).

By the way, CPN and CSE are viable symbols, so I'm curious to know why I'm getting these errors.

Another comment: symbol data quality appears to get worse when I run DTN.IQ at the same time as I do my symbol traversal, and much worse when I have QuoteTracker running at the same time. What I mean, is that I get errors like the above !NONE!, etc, and that data will 'fizzle' out for some reason or another.

This would lead me to believe there are some threading/multi-access issues in the iqfeed code.

I believe that I have fixed this issue. I sent a test build to sruff, and it worked for him. I can email it to you for testing, if you'd like. It could be causing the invalid symbol errors that you are seeing, so why don't you take a look and let me know if it works for you. Thanks!

Natalie

---

Natalie Hannan DTN Market Access, LLC.