Bug #38574 for Math-GSL: Problem with gsl_set_error_handler

Tue Aug 19 00:52:46 2008 sisyphus [...] cpan.org - Ticket created

CC:	jonathan [...] leto.net,thierry.moisan [...] gmail.com
Subject:	Problem with gsl_set_error_handler_off() and static gsl library

If Math::GSL is built against a static gsl library, gsl_set_error_handler_off() doesn't work as expected. Seems that calling gsl_set_error_handler_off() will turn off the error handler for the Math::GSL::Errno package only. As far as all other packages are concerned, the default error handler is still on. The best solution at the moment is to build Math::GSL against a shared gsl library - but if you really must build against a static library then, for some ideas on how to approach this issue, see the "gsl_set_error_handler_off() and static gsl library" thread on math- gsl-dev ( http://groups.google.com/group/math-gsl- dev/t/fdd82288ee21f4f4?hl=en ). One simple but tedious solution mentioned there (with patches for the Histogram, Histogram2D and SF packages) is to create a gsl_set_error_handler_off function for each and every package - so that each package is then capable of turning off the default error handler. OTOH, if you're quite happy to leave the default error handler on, there's no need to take any action at all. Cheers, Rob

Sun Sep 14 20:19:22 2008 jonathan [...] leto.net - Correspondence added

Can someone please test that this is still the case for Math::GSL 0.12 ? Thanks On Tue Aug 19 00:52:46 2008, SISYPHUS wrote: Show quoted text

> If Math::GSL is built against a static gsl library, > gsl_set_error_handler_off() doesn't work as expected. Seems that > calling gsl_set_error_handler_off() will turn off the error handler for > the Math::GSL::Errno package only. As far as all other packages are > concerned, the default error handler is still on. > > The best solution at the moment is to build Math::GSL against a shared > gsl library - but if you really must build against a static library > then, for some ideas on how to approach this issue, see > the "gsl_set_error_handler_off() and static gsl library" thread on math- > gsl-dev ( http://groups.google.com/group/math-gsl- > dev/t/fdd82288ee21f4f4?hl=en ). One simple but tedious solution > mentioned there (with patches for the Histogram, Histogram2D and SF > packages) is to create a gsl_set_error_handler_off function for each > and every package - so that each package is then capable of turning off > the default error handler. > OTOH, if you're quite happy to leave the default error handler on, > there's no need to take any action at all. > > Cheers, > Rob

Sun Sep 14 20:19:23 2008 The RT System itself - Status changed from 'new' to 'open'

Sun Sep 14 20:19:25 2008 jonathan [...] leto.net - Given to LETO

Sun Sep 14 22:51:16 2008 sisyphus1 [...] optusnet.com.au - Correspondence added

CC:	<thierry.moisan [...] gmail.com>
Subject:	Re: [rt.cpan.org #38574] Problem with gsl_set_error_handler_off() and static gsl library
Date:	Mon, 15 Sep 2008 12:49:55 +1000
To:	<bug-Math-Gsl [...] rt.cpan.org>, <sisyphus [...] cpan.org>
From:	"Sisyphus" <sisyphus1 [...] optusnet.com.au>

Show quoted text

----- Original Message ----- From: "Jonathan Leto via RT" <bug-Math-Gsl@rt.cpan.org> To: <sisyphus@cpan.org> Cc: <thierry.moisan@gmail.com> Sent: Monday, September 15, 2008 10:19 AM Subject: [rt.cpan.org #38574] Problem with gsl_set_error_handler_off() and static gsl library

> <URL: http://rt.cpan.org/Ticket/Display.html?id=38574 > > > Can someone please test that this is still the case for Math::GSL 0.12 ? >

Yes, it's still the case for me on Win32. The following errors occur (snipped for brevity): t\Histogram ERROR: cannot find r in cumulative pdf t\Histogram2D ERROR: histogram xrange not in increasing order t\Multiroots ERROR: function incompatible with solver size t\Permutation ERROR: fscanf failed t\SF ERROR: domain error Each of those 5 errors invokes the default error handler. Cheers, Rob

Bug #38574 for Math-GSL: Problem with gsl_set_error_handler_off() and static gsl library