Manlio, so compiling to native machine code works ok but if using ghci byte-code interpreter doesn't ..... can you supply your program please? Vasili On Fri, Jan 9, 2009 at 5:30 PM, Manlio Perillo wrote:

Galchin, Vasili ha scritto:

...

Hi Manlio,

I am the author of this package. Let me think about what you have said.

Regards, Vasili

Thanks.

Note that there are no problems if I compile my program, instead of running it using ghci.

Manlio Perillo

On Sat, Jan 10, 2009 at 5:52 AM, Manlio Perillo wrote:

Galchin, Vasili ha scritto:

...

Manlio,

so compiling to native machine code works ok but if using ghci byte-code interpreter doesn't ..... can you supply your program please?

Right. Can't you reproduce the problem?

The program is very simple (I was just testing your package, since I suggested the use of clock_gettime to Mauricio in a previous post):

import System.Posix.Realtime.RTTime import System.Posix.Realtime.RTDataTypes

main = do time <- clockGetTime Clock_Monotonic; print $ tvSec time print $ tvNsec time

runghc rttime.hs

I suspect that this is a problem with shared library loading in ghci, since the C code you use for your package, is also used by the base package (for the Posix subsystem).

By the way: I don't see reasons to add all that code, since it is not used.

^^^ which code?

However, when I tried to remove all the unused code, executing the program gave me a stack exception (maybe I have removed too many things...).

One personal note: I don't like `tvSec` and `tvNsec`, I think `seconds` and `nanoSeconds` is a better choice. ^^^ ok .. I agree and will change. I asked others for criticisms(constructive) when I put to hackage but didn't get any. This is good ...

Also, it would useful a function to compute elapsed time (maybe a general class in base package, and a specialized instance declaration in posix-realtime for the timespec?)

Vasili

...

[...]

Manlio

On Sat, Jan 10, 2009 at 4:37 PM, Manlio Perillo wrote:

Galchin, Vasili ha scritto:

...

[...] I suspect that this is a problem with shared library loading in ghci, since the C code you use for your package, is also used by the base package (for the Posix subsystem).

By the way: I don't see reasons to add all that code, since it is not used.

^^^ which code?

Code from HsUnix.h and execvpe.h.

...

One personal note: I don't like `tvSec` and `tvNsec`, I think `seconds` and `nanoSeconds` is a better choice. ^^^ ok .. I agree and will change. I asked others for criticisms(constructive) when I put to hackage but didn't get any. This is good ...

I would like to help to develope any wrappers around POSIX API.

^^^ you are suggesting to change current wrapper API? Vasili

Manlio

Hi Manlio, Are you now talking about code in Code from HsUnix.h and execvpe.h? Vasili On Tue, Jan 13, 2009 at 5:09 AM, Manlio Perillo wrote:

Galchin, Vasili ha scritto:

...

[...] I would like to help to develope any wrappers around POSIX API.

^^^ you are suggesting to change current wrapper API?

No, but I don't understand why to link code that seems to be not used.

P.S.: is the problem I have reported riproducible?

I'm on Linux Debian Lenny.

Manlio Perillo

Hi Manlio, ok .. yeh ... I will have to remove the code in HsUnix.h and/or remove references. Currently I am trying to finish another Haskell project. I don't think these include files shouldcause "correctness" problems, yes? If so, I will get to this problem later. ??? Regards, Vasili On Wed, Jan 14, 2009 at 5:03 AM, Manlio Perillo wrote:

Galchin, Vasili ha scritto:

...

Hi Manlio,

Are you now talking about code in Code from HsUnix.h and execvpe.h?

Yes.

Manlio