Problem: error with SMT when generating czmq build stuff

[sappo]

To following error is thrown:

Fatal error in application - aborted
Abort.  Compile with SMT_DETECT for more info.
Aborted (core dumped)

The error occurs the second time ./generate.sh is executed. To reproduce this error do the following:

git clone https://github.com/zeromq/czmq.git
cd czmq
./generate.sh    # Everything works fine
./generate.sh    # Fatal error (see below)

[jschultz]

I tracked the problem down to GSL's script management, and in particular
its cleaning up of scripts to save memory. I don't have a real solution,
but a work-around would be to modify generate.sh to add an argument
'-s:999999999' to the GSL invocation, so that it doesn't try to clean up
the script.

On 02/09/15 20:06, Kevin Sapper wrote:

To following error is thrown:

|Fatal error in application - aborted Abort. Compile with SMT_DETECT for
more info. Aborted (core dumped) |

The error occurs the second time ./generate.sh is executed. To reproduce
this error do the following:

|git clone https://github.com/zeromq/czmq.git cd czmq ./generate.sh #
Everything works fine ./generate.sh # Fatal error (see below) |

—
Reply to this email directly or view it on GitHub
#102.

[sappo]

@jschultz thanks for the workaround, with it the script runs through. Though there is still an problem.
In case of czmq the error with or without workaround will omit the generation of the last 6 lines of a header file in czmq.

[jschultz]

I know plenty about GSL (I wrote most of it) but literally nothing about
czmq. If you can demonstrate that the problem is caused by GSL itself
(which the initial error most certainly was) then I'll take a look at
it, but otherwise I'm handballing this issue to someone else.

On 04/09/15 17:12, Kevin Sapper wrote:

@jschultz https://github.com/jschultz thanks for the workaround, with
it the script runs through. Though there is still an problem.
In case of czmq the error with or without workaround will omit the
generation of the last 6 lines of a header file in czmq.

—
Reply to this email directly or view it on GitHub
#102 (comment).

[sappo]

Could this problem be caused by repeated calls to 'output' without 'close'? Because this seems to solve the issue.

[hintjens]

I've seen this before with gsl, an abort that causes the last part of the
script to not be output. Could be a file buffering issue. Czmq is
irrelevant here, I'm pretty sure, except as a reproducible failure case.
On 4 Sep 2015 13:35, "jschultz" notifications@github.com wrote:

I know plenty about GSL (I wrote most of it) but literally nothing about
czmq. If you can demonstrate that the problem is caused by GSL itself
(which the initial error most certainly was) then I'll take a look at
it, but otherwise I'm handballing this issue to someone else.

On 04/09/15 17:12, Kevin Sapper wrote:

@jschultz https://github.com/jschultz thanks for the workaround, with
it the script runs through. Though there is still an problem.
In case of czmq the error with or without workaround will omit the
generation of the last 6 lines of a header file in czmq.

—
Reply to this email directly or view it on GitHub
#102 (comment).

—
Reply to this email directly or view it on GitHub
#102 (comment).

[jschultz]

Can someone produce a minimal test case that demonstrates the problem
and I'll take a look at it?

On 04/09/15 22:01, Pieter Hintjens wrote:

I've seen this before with gsl, an abort that causes the last part of the
script to not be output. Could be a file buffering issue. Czmq is
irrelevant here, I'm pretty sure, except as a reproducible failure case.
On 4 Sep 2015 13:35, "jschultz" notifications@github.com wrote:

I know plenty about GSL (I wrote most of it) but literally nothing about
czmq. If you can demonstrate that the problem is caused by GSL itself
(which the initial error most certainly was) then I'll take a look at
it, but otherwise I'm handballing this issue to someone else.

On 04/09/15 17:12, Kevin Sapper wrote:

@jschultz https://github.com/jschultz thanks for the workaround, with
it the script runs through. Though there is still an problem.
In case of czmq the error with or without workaround will omit the
generation of the last 6 lines of a header file in czmq.

—
Reply to this email directly or view it on GitHub
#102 (comment).

—
Reply to this email directly or view it on GitHub
#102 (comment).

—
Reply to this email directly or view it on GitHub
#102 (comment).

[sappo]

git clone https://github.com/zeromq/czmq.git
cd czmq
./generate.sh    # Everything works fine
./generate.sh    # Fatal error (see below)

You can find the corrupted file with git status or git diff

[jschultz]

I remember something like this. Didn't we fix this problem? If not, how
did czmq ever work?

On 04/09/15 22:01, Pieter Hintjens wrote:

I've seen this before with gsl, an abort that causes the last part of the
script to not be output. Could be a file buffering issue. Czmq is
irrelevant here, I'm pretty sure, except as a reproducible failure case.
On 4 Sep 2015 13:35, "jschultz" notifications@github.com wrote:

I know plenty about GSL (I wrote most of it) but literally nothing about
czmq. If you can demonstrate that the problem is caused by GSL itself
(which the initial error most certainly was) then I'll take a look at
it, but otherwise I'm handballing this issue to someone else.

On 04/09/15 17:12, Kevin Sapper wrote:

@jschultz https://github.com/jschultz thanks for the workaround, with
it the script runs through. Though there is still an problem.
In case of czmq the error with or without workaround will omit the
generation of the last 6 lines of a header file in czmq.

—
Reply to this email directly or view it on GitHub
#102 (comment).

—
Reply to this email directly or view it on GitHub
#102 (comment).

—
Reply to this email directly or view it on GitHub
#102 (comment).

[hintjens]

So CZMQ is a C library, nothing special. There are a number of different
code generators in there. The one that's failing here is one particular GSL
script. I think there's an exception internally that's causing the output
to be discarded.

On Sat, Sep 5, 2015 at 7:18 AM, jschultz notifications@github.com wrote:

I remember something like this. Didn't we fix this problem? If not, how
did czmq ever work?

On 04/09/15 22:01, Pieter Hintjens wrote:

I've seen this before with gsl, an abort that causes the last part of the
script to not be output. Could be a file buffering issue. Czmq is
irrelevant here, I'm pretty sure, except as a reproducible failure case.
On 4 Sep 2015 13:35, "jschultz" notifications@github.com wrote:

I know plenty about GSL (I wrote most of it) but literally nothing about
czmq. If you can demonstrate that the problem is caused by GSL itself
(which the initial error most certainly was) then I'll take a look at
it, but otherwise I'm handballing this issue to someone else.

On 04/09/15 17:12, Kevin Sapper wrote:

@jschultz https://github.com/jschultz thanks for the workaround,
with
it the script runs through. Though there is still an problem.
In case of czmq the error with or without workaround will omit the
generation of the last 6 lines of a header file in czmq.

—
Reply to this email directly or view it on GitHub
#102 (comment).

—
Reply to this email directly or view it on GitHub
#102 (comment).

—
Reply to this email directly or view it on GitHub
#102 (comment).

—
Reply to this email directly or view it on GitHub
#102 (comment).