segfault depending on code position in source file

older
Advert for: 12 fully-funded 3-year...

Damien Mattei

21 Feb 2019 21 Feb '19

3:13 p.m.

my code is structured like this , in summary: -- this function will return th N°BD from Sidonie for a given name -- note: names have been standardized between Sidonie and WDS getAlphaDelta :: Connection -> String -> IO (Double,Double) getAlphaDelta conn name = do let qry_head_AlphaDelta_AngularDistance = "select alpha,delta from AngularDistance where Nom = ?" :: Query (alpha_delta_rows :: [(Double,Double)]) <- query conn qry_head_AlphaDelta_AngularDistance (Only (name::String)) -- putStrLn $ show alpha_delta_rows return (head alpha_delta_rows) main = do conn <- connect defaultConnectInfo { connectHost = "moita", connectUser = "mattei", connectPassword = "sidonie2", connectDatabase = "sidonie" } -- check getAlphaDelta works: let name3 = "A 7" putStrLn $ show name3 ad3 <- getAlphaDelta conn name3 putStr "ad3 =" putStrLn $ show ad3 this works i got result such as: "A 7" ad3 =(11297.0,-619.0) if i move the 5 lines of code at the end of file it segfault between the begin and end i have some DB connection and queriung , filtering, mapping... any idea? should i really upgrade the version of GHCI? Damien

Attachments:

attachment.html (text/html — 2.8 KB)

Show replies by date

Artem Pelenitsyn

21 Feb 21 Feb

4:34 p.m.

New subject: segfault depending on code position in source file

Damien, It is really hard to tell anything without the ability to reproduce the problem on other systems. Try to come up with some example which is easy to get running AND segfaulting on other computers, so that we could play with it. Maybe post the minimal code in separate github repo and provide the link. -- Best, Artem On Thu, 21 Feb 2019 at 10:14 Damien Mattei wrote:

...

my code is structured like this , in summary:

-- this function will return th N°BD from Sidonie for a given name -- note: names have been standardized between Sidonie and WDS getAlphaDelta :: Connection -> String -> IO (Double,Double) getAlphaDelta conn name = do let qry_head_AlphaDelta_AngularDistance = "select alpha,delta from AngularDistance where Nom = ?" :: Query (alpha_delta_rows :: [(Double,Double)]) <- query conn qry_head_AlphaDelta_AngularDistance (Only (name::String)) -- putStrLn $ show alpha_delta_rows return (head alpha_delta_rows)

main =

do conn <- connect defaultConnectInfo { connectHost = "moita", connectUser = "mattei", connectPassword = "sidonie2", connectDatabase = "sidonie" }

-- check getAlphaDelta works:

let name3 = "A 7" putStrLn $ show name3 ad3 <- getAlphaDelta conn name3 putStr "ad3 =" putStrLn $ show ad3

this works i got result such as: "A 7" ad3 =(11297.0,-619.0)

if i move the 5 lines of code at the end of file it segfault

between the begin and end i have some DB connection and queriung , filtering, mapping...

any idea?

should i really upgrade the version of GHCI?

Damien _______________________________________________ Haskell-Cafe mailing list To (un)subscribe, modify options or view archives go to: http://mail.haskell.org/cgi-bin/mailman/listinfo/haskell-cafe Only members subscribed via the mailman list are allowed to post.

Damien Mattei

6:21 p.m.

New subject: segfault depending on code position in source file

i had attached the whole file, so anyone can read the whole code, perheaps it could help, saying me if there is compilation error on other ghc versions, with mine no error at compilation. The problem to test is that without the database connection which is behind a firewall of course ,the code can not be executed. Me i have no more idea, i will install tomorrow an earlier version of ghc and continue to test.... if it's really a problem i could output the partial result and continue in another language such as Racket or Python... but i'm not at this point and hope to continue with Haskell... Damien On Thu, Feb 21, 2019 at 5:35 PM Artem Pelenitsyn wrote:

...

Damien,

It is really hard to tell anything without the ability to reproduce the problem on other systems. Try to come up with some example which is easy to get running AND segfaulting on other computers, so that we could play with it. Maybe post the minimal code in separate github repo and provide the link.

-- Best, Artem

On Thu, 21 Feb 2019 at 10:14 Damien Mattei wrote:

...
my code is structured like this , in summary:

-- this function will return th N°BD from Sidonie for a given name -- note: names have been standardized between Sidonie and WDS getAlphaDelta :: Connection -> String -> IO (Double,Double) getAlphaDelta conn name = do let qry_head_AlphaDelta_AngularDistance = "select alpha,delta from AngularDistance where Nom = ?" :: Query (alpha_delta_rows :: [(Double,Double)]) <- query conn qry_head_AlphaDelta_AngularDistance (Only (name::String)) -- putStrLn $ show alpha_delta_rows return (head alpha_delta_rows)

main =

do conn <- connect defaultConnectInfo { connectHost = "moita", connectUser = "mattei", connectPassword = "sidonie2", connectDatabase = "sidonie" }

-- check getAlphaDelta works:

let name3 = "A 7" putStrLn $ show name3 ad3 <- getAlphaDelta conn name3 putStr "ad3 =" putStrLn $ show ad3

this works i got result such as: "A 7" ad3 =(11297.0,-619.0)

if i move the 5 lines of code at the end of file it segfault

between the begin and end i have some DB connection and queriung , filtering, mapping...

any idea?

should i really upgrade the version of GHCI?

Damien _______________________________________________ Haskell-Cafe mailing list To (un)subscribe, modify options or view archives go to: http://mail.haskell.org/cgi-bin/mailman/listinfo/haskell-cafe Only members subscribed via the mailman list are allowed to post.

Sven Panne

7:15 p.m.

New subject: segfault depending on code position in source file

Am Do., 21. Feb. 2019 um 19:22 Uhr schrieb Damien Mattei < damien.mattei@gmail.com>:

...

i had attached the whole file, so anyone can read the whole code, perheaps it could help, saying me if there is compilation error on other ghc versions, with mine no error at compilation.

I very much doubt that anyone likes to read >500 lines of Haskell code with lots of external dependencies to help you with your problem. A good hint for Open Source SW in general: If you think you might have found a bug, cut down your example as much as possible, throw out as many external dependencies as you can while still reproducing your problem, then, and only then, post it to other people. You simply can't expect that other people will do that debugging work for you in their spare time. The code you post doesn't really have to make sense, it should just be the minimum amount of code needed to reproduce the problem. In a nutshell: The probability that you get some help is inversely proportional to the size of your example, and even more inversely proportional to the number of external dependencies (libraries etc.).

...

The problem to test is that without the database connection which is behind a firewall of course ,the code can not be executed. [...]

Well, that is basically a show stopper for getting help... I don't want to be mean, quite the opposite: I just want to explain why you probably won't receive any help in the current form. This is not special to the Haskell community, the situation would be very much the same if you posted an equivalent problem of similar size to the Racket/Python/... community. I'm quite sure that most maintainer would be happy to have a look e.g. at a program consisting of a dozen lines without external dependencies.

Damien Mattei

22 Feb 22 Feb

9:55 a.m.

New subject: segfault depending on code position in source file

well i understand the difficulty, i just pos because ,really some times i get incredible help that i coulmdn not even think it possible. i admit the code is and the database connection is a stop help. for the database i cannot do ,for the code i commented a lot and i can resume with this concise code that recreate the problem: it's quite simple i noticed that after one lines (see in code the lines and the upper case comments) the query to DB in getAlphaDelta does not works anymore (as it worked before ) but other functions making queries works that what is strange... {-# LANGUAGE OverloadedStrings #-} {-# LANGUAGE ScopedTypeVariables #-} {-# LANGUAGE FlexibleInstances #-} {-# LANGUAGE MultiParamTypeClasses #-} import Database.MySQL.Simple import Database.MySQL.Simple.QueryResults import Database.MySQL.Simple.Result import Database.MySQL.Simple.QueryParams import Database.MySQL.Simple.Param import Control.Monad import qualified Data.Text as Tx import Debug.Trace import Data.Maybe as Maybe import Text.Read import Data.Tuple.Extra getAlphaDelta :: Connection -> String -> IO (Double,Double) getAlphaDelta conn name = do let qry_head_AlphaDelta_AngularDistance = "select alpha,delta from AngularDistance where Nom = ?" :: Query (alpha_delta_rows :: [(Double,Double)]) <- query conn qry_head_AlphaDelta_AngularDistance (Only (name::String)) return (head alpha_delta_rows) main :: IO () main = do conn <- connect defaultConnectInfo { connectHost = "moita", connectUser = "mattei", connectPassword = "sidonie2", connectDatabase = "sidonie" } -- check getAlphaDelta works: -- HERE getAlphaDelta WORKS let name = "WOR 7" putStrLn $ show name ad <- getAlphaDelta conn name putStr "ad =" putStrLn $ show ad (names ::[Only Tx.Text])<- query_ conn "SELECT Nom FROM AngularDistance WHERE distance > 0.000278" -- AFTER THE PREVIOUS LINE getAlphaDelta WILL NOT WORKS AND ISSUE A SEGFAULT IN GHCI let name3 = "A 7" putStrLn $ show name3 ad3 <- getAlphaDelta conn name3 putStr "ad3 =" putStrLn $ show ad3 close conn here is the output: [mattei@localhost Haskell]$ ghci GHCi, version 8.2.2: http://www.haskell.org/ghc/ :? for help Prelude> :load UpdateSidonie.hs [1 of 1] Compiling Main ( UpdateSidonie.hs, interpreted ) Ok, one module loaded. *Main> main "WOR 7" ad =(17386.0,7120.0) "A 7" Segmentation fault (core dumped) as you see the first call to getAlphaDelta is ok but the second issue a Segmentation fault. i will try now with ghci 8.4 .... and hiope to give you some news... Damien On Thu, Feb 21, 2019 at 8:15 PM Sven Panne wrote:

...

Am Do., 21. Feb. 2019 um 19:22 Uhr schrieb Damien Mattei < damien.mattei@gmail.com>:

...
i had attached the whole file, so anyone can read the whole code, perheaps it could help, saying me if there is compilation error on other ghc versions, with mine no error at compilation.

I very much doubt that anyone likes to read >500 lines of Haskell code with lots of external dependencies to help you with your problem. A good hint for Open Source SW in general: If you think you might have found a bug, cut down your example as much as possible, throw out as many external dependencies as you can while still reproducing your problem, then, and only then, post it to other people. You simply can't expect that other people will do that debugging work for you in their spare time. The code you post doesn't really have to make sense, it should just be the minimum amount of code needed to reproduce the problem.

In a nutshell: The probability that you get some help is inversely proportional to the size of your example, and even more inversely proportional to the number of external dependencies (libraries etc.).

...
The problem to test is that without the database connection which is behind a firewall of course ,the code can not be executed. [...]

Well, that is basically a show stopper for getting help...

I don't want to be mean, quite the opposite: I just want to explain why you probably won't receive any help in the current form. This is not special to the Haskell community, the situation would be very much the same if you posted an equivalent problem of similar size to the Racket/Python/... community. I'm quite sure that most maintainer would be happy to have a look e.g. at a program consisting of a dozen lines without external dependencies.

Damien Mattei

10:39 a.m.

New subject: segfault depending on code position in source file

well , good news :-) i tested the same code on a 64 bits Linux Centos 7.2 with GHC 8.4 and it works fine... so the problem is only under 32bits Linux Fedora Core 29 with GHC 8.2 we had a system failure or security issue last month and i had to upgrade the 32 bit CPU 3 weeks ago to latest Linux Fedora Core version but this had the side effect to downgrade the Haskell GHC compiler to 8.2 version and the 8.2 version or 32 bit platform have bugs if i can i will try to test with the 8.4 GHC on the 32 bit platform to see if the problem was with 32bit vs 64bit; possible difference with Fedora and CentOS also, but i think it's the 8.2 compiler that cause the bug... regards, damien On Fri, Feb 22, 2019 at 10:55 AM Damien Mattei wrote:

...

well i understand the difficulty, i just pos because ,really some times i get incredible help that i coulmdn not even think it possible. i admit the code is and the database connection is a stop help. for the database i cannot do ,for the code i commented a lot and i can resume with this concise code that recreate the problem:

it's quite simple i noticed that after one lines (see in code the lines and the upper case comments) the query to DB in getAlphaDelta does not works anymore (as it worked before ) but other functions making queries works that what is strange...

{-# LANGUAGE OverloadedStrings #-} {-# LANGUAGE ScopedTypeVariables #-} {-# LANGUAGE FlexibleInstances #-} {-# LANGUAGE MultiParamTypeClasses #-} import Database.MySQL.Simple import Database.MySQL.Simple.QueryResults import Database.MySQL.Simple.Result

import Database.MySQL.Simple.QueryParams import Database.MySQL.Simple.Param

import Control.Monad import qualified Data.Text as Tx import Debug.Trace import Data.Maybe as Maybe import Text.Read import Data.Tuple.Extra

getAlphaDelta :: Connection -> String -> IO (Double,Double) getAlphaDelta conn name = do let qry_head_AlphaDelta_AngularDistance = "select alpha,delta from AngularDistance where Nom = ?" :: Query (alpha_delta_rows :: [(Double,Double)]) <- query conn qry_head_AlphaDelta_AngularDistance (Only (name::String)) return (head alpha_delta_rows)

main :: IO ()

main =

do conn <- connect defaultConnectInfo { connectHost = "moita", connectUser = "mattei", connectPassword = "sidonie2", connectDatabase = "sidonie" }

-- check getAlphaDelta works: -- HERE getAlphaDelta WORKS let name = "WOR 7" putStrLn $ show name ad <- getAlphaDelta conn name putStr "ad =" putStrLn $ show ad

(names ::[Only Tx.Text])<- query_ conn "SELECT Nom FROM AngularDistance WHERE distance > 0.000278"

-- AFTER THE PREVIOUS LINE getAlphaDelta WILL NOT WORKS AND ISSUE A SEGFAULT IN GHCI let name3 = "A 7" putStrLn $ show name3 ad3 <- getAlphaDelta conn name3 putStr "ad3 =" putStrLn $ show ad3

close conn

here is the output:

[mattei@localhost Haskell]$ ghci GHCi, version 8.2.2: http://www.haskell.org/ghc/ :? for help Prelude> :load UpdateSidonie.hs [1 of 1] Compiling Main ( UpdateSidonie.hs, interpreted ) Ok, one module loaded. *Main> main "WOR 7" ad =(17386.0,7120.0) "A 7" Segmentation fault (core dumped)

as you see the first call to getAlphaDelta is ok but the second issue a Segmentation fault.

i will try now with ghci 8.4 .... and hiope to give you some news...

Damien

On Thu, Feb 21, 2019 at 8:15 PM Sven Panne wrote:

...
Am Do., 21. Feb. 2019 um 19:22 Uhr schrieb Damien Mattei < damien.mattei@gmail.com>:

...
i had attached the whole file, so anyone can read the whole code, perheaps it could help, saying me if there is compilation error on other ghc versions, with mine no error at compilation.

I very much doubt that anyone likes to read >500 lines of Haskell code with lots of external dependencies to help you with your problem. A good hint for Open Source SW in general: If you think you might have found a bug, cut down your example as much as possible, throw out as many external dependencies as you can while still reproducing your problem, then, and only then, post it to other people. You simply can't expect that other people will do that debugging work for you in their spare time. The code you post doesn't really have to make sense, it should just be the minimum amount of code needed to reproduce the problem.

In a nutshell: The probability that you get some help is inversely proportional to the size of your example, and even more inversely proportional to the number of external dependencies (libraries etc.).

...
The problem to test is that without the database connection which is behind a firewall of course ,the code can not be executed. [...]

Well, that is basically a show stopper for getting help...

I don't want to be mean, quite the opposite: I just want to explain why you probably won't receive any help in the current form. This is not special to the Haskell community, the situation would be very much the same if you posted an equivalent problem of similar size to the Racket/Python/... community. I'm quite sure that most maintainer would be happy to have a look e.g. at a program consisting of a dozen lines without external dependencies.

Sven Panne

11:35 a.m.

New subject: segfault depending on code position in source file

Am Fr., 22. Feb. 2019 um 11:39 Uhr schrieb Damien Mattei < damien.mattei@gmail.com>:

...

[...] but i think it's the 8.2 compiler that cause the bug...

After a quick look at your code, I very much doubt that. You use the mysql-simple package, which in turn uses the mysql package, which in turn uses the native mysqlclient library. My (and probably most people's) reaction here is: It is much, much more probable that either a) you use the package/library incorrectly or b) the package/library has a bug than that you have discovered a compiler bug for otherwise farily trivial code. Your best bet is to either rip out the mysql dependencies and still reproduce the bug or issue a bug report at e.g. https://github.com/paul-rouse/mysql-simple/issues. Note that I'm not saying that GHC is bug-free, it is just a question where people will probably like to spend their time on. Given the fact that there are tons of buggy packages/libraries with questionable quality out there, it is basically *your* onus to show that your problem is GHC's fault. Just my 2c...

...

Damien Mattei

2:09 p.m.

New subject: segfault depending on code position in source file

it's not me that say first ghc can have bug. There is a misunderstanding, i first thought that if there was a problem it was not the core ghc interpreter/compiler (note that i never try to compile with ghc, possible the behavior is different in interpreter than in compile mode) but i had a lot of doubt about the stability of the mysql* packages. BUT i just made more test with the same mysql* package (version is the latest,just installed with cabal) on another system this time 32bits: [mattei@asteroide ~]$ uname -a Linux asteroide 3.10.0-862.3.3.el7.x86_64 #1 SMP Fri Jun 15 04:15:27 UTC 2018 x86_64 x86_64 x86_64 GNU/Linux [mattei@asteroide ~]$ cat /etc/redhat-release CentOS Linux release 7.5.1804 (Core) [mattei@asteroide ~]$ ghci GHCi, version 8.4.3: http://www.haskell.org/ghc/ :? for help ghci version 8.4.3 and the program works fine, this does not means that the ghci v8.2 have bugs nor the mysql* packages , just means that there is some incompatibilies between ghc 8.2 and the latest versions of mysql* packages (mysql-simple-0.4.5) i think there is something that can be enhanced in the cabal or the package installion script of mysql* packages to detect those incompatibilities but i do not know how, perheaps if with cabal and ghc 8.2 i had installed an earlier verison of mysql* it could have worked without problem. Damien On Fri, Feb 22, 2019 at 12:35 PM Sven Panne wrote:

...

Am Fr., 22. Feb. 2019 um 11:39 Uhr schrieb Damien Mattei < damien.mattei@gmail.com>:

...
[...] but i think it's the 8.2 compiler that cause the bug...

After a quick look at your code, I very much doubt that. You use the mysql-simple package, which in turn uses the mysql package, which in turn uses the native mysqlclient library. My (and probably most people's) reaction here is: It is much, much more probable that either

a) you use the package/library incorrectly

or

b) the package/library has a bug

than that you have discovered a compiler bug for otherwise farily trivial code. Your best bet is to either rip out the mysql dependencies and still reproduce the bug or issue a bug report at e.g. https://github.com/paul-rouse/mysql-simple/issues.

Note that I'm not saying that GHC is bug-free, it is just a question where people will probably like to spend their time on. Given the fact that there are tons of buggy packages/libraries with questionable quality out there, it is basically *your* onus to show that your problem is GHC's fault.

Just my 2c...

...

Vanessa McHale

21 Feb 21 Feb

4:41 p.m.

New subject: segfault depending on code position in source file

Well, it's kind of difficult to tell if it's a compiler bug without at least trying different GHC versions. As it stands I don't think there's enough information to diagnose the problem. Perhaps provide something reproducible that we could build on our machines? Cheers, Vanessa McHale On 2/21/19 9:13 AM, Damien Mattei wrote:

...

my code is structured like this , in summary:

-- this function will return th N°BD from Sidonie for a given name -- note: names have been standardized between Sidonie and WDS getAlphaDelta :: Connection -> String -> IO (Double,Double) getAlphaDelta conn name = do             let qry_head_AlphaDelta_AngularDistance = "select alpha,delta from AngularDistance where Nom = ?" :: Query             (alpha_delta_rows :: [(Double,Double)]) <- query conn qry_head_AlphaDelta_AngularDistance (Only (name::String)) --            putStrLn $ show alpha_delta_rows             return (head alpha_delta_rows)

main =

do     conn <- connect defaultConnectInfo       { connectHost = "moita",         connectUser = "mattei",         connectPassword = "sidonie2",         connectDatabase = "sidonie" }

-- check getAlphaDelta works:     let name3 = "A     7"     putStrLn $ show name3     ad3 <- getAlphaDelta conn name3     putStr "ad3 ="     putStrLn $ show ad3

this works i got result such as: "A     7" ad3 =(11297.0,-619.0)

if i move the 5 lines of code at the end of file it segfault

between the begin and end i have some DB connection and queriung , filtering, mapping...

any idea?

should i really upgrade the version of GHCI?

Damien

_______________________________________________ Haskell-Cafe mailing list To (un)subscribe, modify options or view archives go to: http://mail.haskell.org/cgi-bin/mailman/listinfo/haskell-cafe Only members subscribed via the mailman list are allowed to post.

Niklas Hambüchen

5:48 p.m.

New subject: segfault depending on code position in source file

...

should i really upgrade the version of GHCI?

I don't follow. Do you get a segfault when interpreting your code in ghci, or when compiling your program to a binary with ghc? And on what platform? If you get a segfault, it is often worth running your program under gdb, and check in the crash backtrace whether it crashed from Haskell code or a C foreign function call. Niklas

Niklas Hambüchen

5:55 p.m.

New subject: segfault depending on code position in source file

On 21/02/2019 6:48 PM, Niklas Hambüchen wrote:

...

Do you get a segfault when interpreting your code in ghci, or when compiling your program to a binary with ghc?

Ah, I see now that there was another email thread where you said it was in ghci, but you didn't to that one in your email so I missed it. I guess my recommendations stands: Try attach to your ghci with gdb and check whether you get a C location for your crash.

2327

Age (days ago)

2328

Last active (days ago)

List overview

Download

10 comments

5 participants

participants (5)

Artem Pelenitsyn
Damien Mattei
Niklas Hambüchen
Sven Panne
Vanessa McHale