Bug#545806: libgsl0-dev: GSL should deal with an uninitialised rng better
Brian, On 11 September 2009 at 16:54, Brian Gough wrote: | At Thu, 10 Sep 2009 07:55:25 -0500, | Dirk Eddelbuettel wrote: | > Tough one. SEGV is bad. On the other hand, the GSL RMGs came in a 'framework' | > so there may be numerous entry points. | > | > Brian, any thougths? | | Hello, | | By convention we've never checked whether pointers passed in by the | user are actually non-NULL. That was my gut feel. Thanks for confirming this. | | This is mainly due to the amount of work in adding these checks | everywhere. | | Also there would be a somewhat arbitrary distinction between functions | where we would check, and those where we wouldn't because they would | add too much overhead. Matthew: Ok if I close this bug report 'as it is a feature' ? Dirk | | -- | Brian Gough | | Network Theory Ltd, | Publishing Free Software Manuals --- http://www.network-theory.co.uk/ | | -- Three out of two people have difficulties with fractions. -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org
Bug#545806: libgsl0-dev: GSL should deal with an uninitialised rng better
At Thu, 10 Sep 2009 07:55:25 -0500, Dirk Eddelbuettel wrote: > Tough one. SEGV is bad. On the other hand, the GSL RMGs came in a 'framework' > so there may be numerous entry points. > > Brian, any thougths? Hello, By convention we've never checked whether pointers passed in by the user are actually non-NULL. This is mainly due to the amount of work in adding these checks everywhere. Also there would be a somewhat arbitrary distinction between functions where we would check, and those where we wouldn't because they would add too much overhead. -- Brian Gough Network Theory Ltd, Publishing Free Software Manuals --- http://www.network-theory.co.uk/ -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org
Bug#545806: libgsl0-dev: GSL should deal with an uninitialised rng better
On 10 September 2009 at 12:48, Matthew Vernon wrote: | Dirk Eddelbuettel wrote: | | | | > Or do you claim that not-using | > | >gsl_rng_alloc (gsl_rng_default) | > | > leading to a SEGV is a invalid behaviour? I am much less convinced that that | > is a valid bug report -- it's much like using un-alloc'ed pointer types in | > my book. | | That is my claim, yes; I'd content that a very simple | sanity-check-and-error message would be better behaviour from a library | in this case. I conceed that it's arguable, but IMAO gsl should emit an | error rather than SEGV on bad input to any of its functions, unless it's | unreasonably expensive to do the error-checking. Tough one. SEGV is bad. On the other hand, the GSL RMGs came in a 'framework' so there may be numerous entry points. Brian, any thougths? (Oh, and I missed gsl 1.13 by a day but it is in Debian unstable as of yesterday eve.) Dirk | | Regards, | | Matthew | | -- | Matthew Vernon, Research Fellow | Ecology and Epidemiology Group, | University of Warwick | http://blogs.warwick.ac.uk/mcvernon -- Three out of two people have difficulties with fractions. -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org
Bug#545806: libgsl0-dev: GSL should deal with an uninitialised rng better
Dirk Eddelbuettel wrote: Or do you claim that not-using gsl_rng_alloc (gsl_rng_default) leading to a SEGV is a invalid behaviour? I am much less convinced that that is a valid bug report -- it's much like using un-alloc'ed pointer types in my book. That is my claim, yes; I'd content that a very simple sanity-check-and-error message would be better behaviour from a library in this case. I conceed that it's arguable, but IMAO gsl should emit an error rather than SEGV on bad input to any of its functions, unless it's unreasonably expensive to do the error-checking. Regards, Matthew -- Matthew Vernon, Research Fellow Ecology and Epidemiology Group, University of Warwick http://blogs.warwick.ac.uk/mcvernon -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org
Bug#545806: libgsl0-dev: GSL should deal with an uninitialised rng better
On 9 September 2009 at 10:27, Matthew Vernon wrote:
| Hi,
|
| If you call any of the random-number functions in GSL and forget to
| initialise the random number generator (or initialise it wrongly), GSL
| SEGVs. This is poor - it should produce an error message. I deal with
| queries from people this has confused surprisingly frequently.
Can you please supply a reproducible example? If I bastardize an old example
of mine and comment-out 'gsl_rng_env_setup()' it still runs fine as shown below.
e...@ron:/tmp> cat test_gsl_rng.c
/* minimal gsl program from the info pages */
#include
#include
gsl_rng * r ; /* global generator */
int
main ()
{
/* gsl_rng_env_setup() ; */
r = gsl_rng_alloc (gsl_rng_default);
printf("generator type: %s\n", gsl_rng_name (r));
printf("seed = %u\n", gsl_rng_default_seed);
printf("first value = %u\n", gsl_rng_get (r)) ;
}
/*
* Local variables:
* compile-command: "gcc -W test_gsl_rng.c -o test_gsl_rng -lblas -lgsl -lm"
* End:
*/
e...@ron:/tmp> gcc -W test_gsl_rng.c -o test_gsl_rng -lblas -lgsl -lm
e...@ron:/tmp> ./test_gsl_rng
generator type: mt19937
seed = 0
first value = 4293858116
e...@ron:/tmp>
Or do you claim that not-using
gsl_rng_alloc (gsl_rng_default)
leading to a SEGV is a invalid behaviour? I am much less convinced that that
is a valid bug report -- it's much like using un-alloc'ed pointer types in
my book.
Dirk
|
| Regards,
|
| Matthew
|
| -- System Information:
| Debian Release: 5.0.3
| APT prefers stable
| APT policy: (500, 'stable')
| Architecture: amd64 (x86_64)
|
| Kernel: Linux 2.6.20-2-macpro-amd64 (SMP w/4 CPU cores)
| Locale: LANG=en_GB.UTF-8, LC_CTYPE=en_GB.UTF-8 (charmap=UTF-8)
| Shell: /bin/sh linked to /bin/bash
|
| Versions of packages libgsl0-dev depends on:
| ii libgsl0ldbl 1.11+dfsg-1 GNU Scientific Library (GSL) --
li
|
| libgsl0-dev recommends no packages.
|
| libgsl0-dev suggests no packages.
|
| -- no debconf information
|
|
--
Three out of two people have difficulties with fractions.
--
To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org
Bug#545806: libgsl0-dev: GSL should deal with an uninitialised rng better
Package: libgsl0-dev Version: 1.11+dfsg-1 Severity: normal Hi, If you call any of the random-number functions in GSL and forget to initialise the random number generator (or initialise it wrongly), GSL SEGVs. This is poor - it should produce an error message. I deal with queries from people this has confused surprisingly frequently. Regards, Matthew -- System Information: Debian Release: 5.0.3 APT prefers stable APT policy: (500, 'stable') Architecture: amd64 (x86_64) Kernel: Linux 2.6.20-2-macpro-amd64 (SMP w/4 CPU cores) Locale: LANG=en_GB.UTF-8, LC_CTYPE=en_GB.UTF-8 (charmap=UTF-8) Shell: /bin/sh linked to /bin/bash Versions of packages libgsl0-dev depends on: ii libgsl0ldbl 1.11+dfsg-1 GNU Scientific Library (GSL) -- li libgsl0-dev recommends no packages. libgsl0-dev suggests no packages. -- no debconf information -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org
