Having a very annoying problem related with (I think) lazy IO

older
Typesafe multi-table PostgreSQL...

Ömer Sinan Ağacan

12 Dec 2014 12 Dec '14

3:58 p.m.

Hi all, I'm having a horrible bug in my program: a print statement that doesn't force any values(e.g. it just prints a static string) effecting the behavior of the program. I tried hard to produce a minimal example, but couldn't manage it. So instead I'm just going to point to the source code of my program: https://github.com/osa1/rho-torrent/blob/master/src/Rho/Listener.hs#L125 If I remove this print statement everything goes weird. I think when I remove this statement some effects are not happening, or happening in wrong times. Here's how I'm testing the different behavior: This is a network application and I'm creating an environment in my system and making this program talk to another program. I run my program 10 times without the print statement, and run 10 times with it. Both groups produce same output in every time, but outputs of first group and second group are different from each other. (apart from extra prints, of course) So this print statement effects some other things, and expected behavior is the one happening with print statement. I know the program is big for anyone to figure things in short amount of time, I'm just looking for tips and pitfalls that I may be falling. Does anyone have any ideas what I may be doing wrong? Note that the program doesn't use any unsafe stuff. Thanks in advance.

Show replies by date

Yuras Shumovich

12 Dec 12 Dec

4:20 p.m.

New subject: Having a very annoying problem related with (I think) lazy IO

What if you replace the `putStrLn` with `yield` or `threadDelay 1`? Probably the loop is too fast, and it doesn't give other parts of your program to take a lock. On Fri, 2014-12-12 at 17:58 +0200, Ömer Sinan Ağacan wrote:

...

Hi all,

I'm having a horrible bug in my program: a print statement that doesn't force any values(e.g. it just prints a static string) effecting the behavior of the program.

I tried hard to produce a minimal example, but couldn't manage it. So instead I'm just going to point to the source code of my program: https://github.com/osa1/rho-torrent/blob/master/src/Rho/Listener.hs#L125

If I remove this print statement everything goes weird. I think when I remove this statement some effects are not happening, or happening in wrong times.

Here's how I'm testing the different behavior: This is a network application and I'm creating an environment in my system and making this program talk to another program.

I run my program 10 times without the print statement, and run 10 times with it.

Both groups produce same output in every time, but outputs of first group and second group are different from each other. (apart from extra prints, of course)

So this print statement effects some other things, and expected behavior is the one happening with print statement.

I know the program is big for anyone to figure things in short amount of time, I'm just looking for tips and pitfalls that I may be falling. Does anyone have any ideas what I may be doing wrong? Note that the program doesn't use any unsafe stuff.

Thanks in advance. _______________________________________________ Haskell-Cafe mailing list Haskell-Cafe@haskell.org http://www.haskell.org/mailman/listinfo/haskell-cafe

Ömer Sinan Ağacan

4:33 p.m.

New subject: Having a very annoying problem related with (I think) lazy IO

...

Probably the loop is too fast, and it doesn't give other parts of your

Interesting, I think that worked. (I'll try on bigger examples and update again) program to take a lock. Are there any ways to test for this? How do I know if this is really the case? 2014-12-12 18:20 GMT+02:00 Yuras Shumovich :

...

What if you replace the `putStrLn` with `yield` or `threadDelay 1`? Probably the loop is too fast, and it doesn't give other parts of your program to take a lock.

On Fri, 2014-12-12 at 17:58 +0200, Ömer Sinan Ağacan wrote:

...
Hi all,

I'm having a horrible bug in my program: a print statement that doesn't force any values(e.g. it just prints a static string) effecting the behavior of the program.

I tried hard to produce a minimal example, but couldn't manage it. So instead I'm just going to point to the source code of my program: https://github.com/osa1/rho-torrent/blob/master/src/Rho/Listener.hs#L125

If I remove this print statement everything goes weird. I think when I remove this statement some effects are not happening, or happening in wrong times.

Here's how I'm testing the different behavior: This is a network application and I'm creating an environment in my system and making this program talk to another program.

I run my program 10 times without the print statement, and run 10 times with it.

Both groups produce same output in every time, but outputs of first group and second group are different from each other. (apart from extra prints, of course)

So this print statement effects some other things, and expected behavior is the one happening with print statement.

I know the program is big for anyone to figure things in short amount of time, I'm just looking for tips and pitfalls that I may be falling. Does anyone have any ideas what I may be doing wrong? Note that the program doesn't use any unsafe stuff.

Thanks in advance. _______________________________________________ Haskell-Cafe mailing list Haskell-Cafe@haskell.org http://www.haskell.org/mailman/listinfo/haskell-cafe

Ömer Sinan Ağacan

4:38 p.m.

New subject: Having a very annoying problem related with (I think) lazy IO

Amazing, that really worked. I'm still confused about why this is happening though.

...

it doesn't give other parts of your program to take a lock.

What lock are we talking about here? 2014-12-12 18:33 GMT+02:00 Ömer Sinan Ağacan :

...

Interesting, I think that worked. (I'll try on bigger examples and update again)

...
Probably the loop is too fast, and it doesn't give other parts of your program to take a lock.

Are there any ways to test for this? How do I know if this is really the case?

2014-12-12 18:20 GMT+02:00 Yuras Shumovich :

...
What if you replace the `putStrLn` with `yield` or `threadDelay 1`? Probably the loop is too fast, and it doesn't give other parts of your program to take a lock.

On Fri, 2014-12-12 at 17:58 +0200, Ömer Sinan Ağacan wrote:

...
Hi all,

I'm having a horrible bug in my program: a print statement that doesn't force any values(e.g. it just prints a static string) effecting the behavior of the program.

I tried hard to produce a minimal example, but couldn't manage it. So instead I'm just going to point to the source code of my program: https://github.com/osa1/rho-torrent/blob/master/src/Rho/Listener.hs#L125

If I remove this print statement everything goes weird. I think when I remove this statement some effects are not happening, or happening in wrong times.

Here's how I'm testing the different behavior: This is a network application and I'm creating an environment in my system and making this program talk to another program.

I run my program 10 times without the print statement, and run 10 times with it.

Both groups produce same output in every time, but outputs of first group and second group are different from each other. (apart from extra prints, of course)

So this print statement effects some other things, and expected behavior is the one happening with print statement.

I know the program is big for anyone to figure things in short amount of time, I'm just looking for tips and pitfalls that I may be falling. Does anyone have any ideas what I may be doing wrong? Note that the program doesn't use any unsafe stuff.

Thanks in advance. _______________________________________________ Haskell-Cafe mailing list Haskell-Cafe@haskell.org http://www.haskell.org/mailman/listinfo/haskell-cafe

Yuras Shumovich

4:53 p.m.

New subject: Having a very annoying problem related with (I think) lazy IO

Well, I'm not sure, I just wrote what my nose suggested me. I can't count how many times I got something similar in my own code, and each time it was a thread that runs too fast. I mean the `lock` at line 121 (`takeMVar lock`). I don't know whether it is used somewhere else though. It also could be `updated` MVar, or it is a flaw in ghc's scheduler... You can try ThreadScope (see e.g. https://www.haskell.org/haskellwiki/ThreadScope ) to see a difference between two versions (with and without `putStrLn`). It is hard to suggest anything specific without deep study. On Fri, 2014-12-12 at 18:38 +0200, Ömer Sinan Ağacan wrote:

...

Amazing, that really worked. I'm still confused about why this is happening though.

...
it doesn't give other parts of your program to take a lock.

What lock are we talking about here?

2014-12-12 18:33 GMT+02:00 Ömer Sinan Ağacan :

...
Interesting, I think that worked. (I'll try on bigger examples and update again)

...
Probably the loop is too fast, and it doesn't give other parts of your program to take a lock.

Are there any ways to test for this? How do I know if this is really the case?

2014-12-12 18:20 GMT+02:00 Yuras Shumovich :

...
What if you replace the `putStrLn` with `yield` or `threadDelay 1`? Probably the loop is too fast, and it doesn't give other parts of your program to take a lock.

On Fri, 2014-12-12 at 17:58 +0200, Ömer Sinan Ağacan wrote:

...
Hi all,

I'm having a horrible bug in my program: a print statement that doesn't force any values(e.g. it just prints a static string) effecting the behavior of the program.

I tried hard to produce a minimal example, but couldn't manage it. So instead I'm just going to point to the source code of my program: https://github.com/osa1/rho-torrent/blob/master/src/Rho/Listener.hs#L125

If I remove this print statement everything goes weird. I think when I remove this statement some effects are not happening, or happening in wrong times.

Here's how I'm testing the different behavior: This is a network application and I'm creating an environment in my system and making this program talk to another program.

I run my program 10 times without the print statement, and run 10 times with it.

Both groups produce same output in every time, but outputs of first group and second group are different from each other. (apart from extra prints, of course)

So this print statement effects some other things, and expected behavior is the one happening with print statement.

I know the program is big for anyone to figure things in short amount of time, I'm just looking for tips and pitfalls that I may be falling. Does anyone have any ideas what I may be doing wrong? Note that the program doesn't use any unsafe stuff.

Thanks in advance. _______________________________________________ Haskell-Cafe mailing list Haskell-Cafe@haskell.org http://www.haskell.org/mailman/listinfo/haskell-cafe

3860

Age (days ago)

3860

Last active (days ago)

List overview

Download

4 comments

2 participants

participants (2)

Yuras Shumovich
Ömer Sinan Ağacan