Bug#799036: [Pkg-julia-devel] Bug#799036: openspecfun: please build everywhere

[Graham Inggs]

Hi Sébastien

On 16 September 2015 at 22:47, Sébastien Villemot  wrote:
> It builds, but I am not sure it works.

I wrote the attached test program and built openspecfun on amd64,
armhf and ppc64el.
All versions output the same results.

> There is some low level stuff in
> the rem_pio2/ directory. For example, in rem_pio2/fpmath.h, there is
> this comment: "Currently assumes Intel platform", followed by an
> alternative that seems to assume either amd64 or i386.

Note that if the platform is neither x86_64 nor i386, then neither
amd64_fpmath.h nor i386_fpmath.h is included.
In fact, the things that are defined in these headers don't seem to be
used anywhere in rem_pio2, and removing them didn't seem to have any
effect on the results.

Please consider again allowing builds on all architectures.

Regards
Graham
#include 
#include 
#include "../openspecfun-0.4/rem_pio2/math_private.h"

void test__ieee754_rem_pio2(void)
{
	double pi, x, y[2];
	int i;

	printf("__ieee754_rem_pio2()\n");
	pi = 4 * atan(1);
	for(i = 0; i < 17; i++) {
		x = 2 * pi * i / 16;
		__ieee754_rem_pio2(x, y);
		printf("x=%f, y[0]=%f, y[1]=%f\n", x, y[0], y[1]);
	}
}

void test__ieee754_rem_pio2f(void)
{
	double pi, x, y;
	int i;

	printf("__ieee754_rem_pio2f()\n");
	pi = 4 * atan(1);
	for(i = 0; i < 17; i++) {
		x = 2 * pi * i / 16;
		__ieee754_rem_pio2f(x, );
		printf("x=%f, y=%f\n", x, y);
	}
}

void test__kernel_rem_pio2(void)
{
	double pi, x[1], y[4];
	int i, e0, nx, prec;

	printf("__kernel_rem_pio2()\n");
	pi = 4 * atan(1);
	e0 = 0;
	nx = 1;
	prec = 1;
	for(i = 1; i < 17; i++) {
		x[0] = 2 * pi * i / 16;
		__kernel_rem_pio2(x, y, e0, nx, prec);
		printf("x[0]=%f, y[0]=%f\n", x[0], y[0]);
	}
}

void main(void)
{
	test__ieee754_rem_pio2();
	test__ieee754_rem_pio2f();
	test__kernel_rem_pio2();
}

Bug#799036: openspecfun: please build everywhere

[Jonas Smedegaard]

Benefit of targeting any arch instead of known succesful ones is to help 
porters identify where work is needed - a.k.a. not hide problems, as 
we're all abiding to in our Social Contract.

You need not be concerned about FTBFS on strange archs affecting the 
package migrating to testing: Only FTBFS on archs already in testing are 
treated as severe - i.e. only regressions count.

Only if this package lacks a proper testsuite and you know that builds 
on some archs even though technically succeeding will produce broken 
code.  If that's the case I urge you to make it more visible e.g. to 
porters - one way could be to add arch checks in debian/rules that 
checks for build architecture and fails with a suitably descriptive 
error message that porters can see in build logs for those archs.

Hope that helps,

 - Jonas

-- 
 * Jonas Smedegaard - idealist & Internet-arkitekt
 * Tlf.: +45 40843136  Website: http://dr.jones.dk/

 [x] quote me freely  [ ] ask before reusing  [ ] keep private


signature.asc
Description: signature

Bug#799036: [Pkg-julia-devel] Bug#799036: Bug#799036: openspecfun: please build everywhere

[Graham Inggs]

On 17/09/2015 17:38, Sébastien Villemot wrote:

I don't understand why this would be better. It seems to me that it
would just eat buildd resources, with no clear benefit… Or am I missing
something?


In my opinion, restricting architectures is hiding the problem.

In the case of openspecfun, it builds in seconds (even on ARM) so 
wouldn't be much of a burden on the buildds.


If the build logs show a failure, it may attract the attention of 
someone with an interest in that architecture and the problem could get 
fixed.

Bug#799036: [Pkg-julia-devel] Bug#799036: Bug#799036: Bug#799036: openspecfun: please build everywhere

[Viral Shah]

I would be ok with that too. The rem_pio2 is just thrown in there for 
convenience - it should be removed eventually.

-viral



> On 17-Sep-2015, at 10:06 pm, Graham Inggs  wrote:
> 
> On 17/09/2015 17:38, Sébastien Villemot wrote:
>> I don't understand why this would be better. It seems to me that it
>> would just eat buildd resources, with no clear benefit… Or am I missing
>> something?
> 
> In my opinion, restricting architectures is hiding the problem.
> 
> In the case of openspecfun, it builds in seconds (even on ARM) so wouldn't be 
> much of a burden on the buildds.
> 
> If the build logs show a failure, it may attract the attention of someone 
> with an interest in that architecture and the problem could get fixed.
> 
> ___
> Pkg-julia-devel mailing list
> pkg-julia-de...@lists.alioth.debian.org
> http://lists.alioth.debian.org/cgi-bin/mailman/listinfo/pkg-julia-devel

Bug#799036: [Pkg-julia-devel] Bug#799036: Bug#799036: openspecfun: please build everywhere

[Sébastien Villemot]

Le jeudi 17 septembre 2015 à 17:24 +0200, Graham Inggs a écrit :

> I think it would be better if the archs were not restricted and the 
> package failed on the buildds.

I don't understand why this would be better. It seems to me that it
would just eat buildd resources, with no clear benefit… Or am I missing
something?

-- 
 .''`.Sébastien Villemot
: :' :Debian Developer
`. `' http://sebastien.villemot.name
  `-  GPG Key: 4096R/381A7594

Bug#799036: [Pkg-julia-devel] Bug#799036: openspecfun: please build everywhere

[Graham Inggs]

Hi Sébastien

On 16/09/2015 22:47, Sébastien Villemot wrote:

It builds, but I am not sure it works. There is some low level stuff in
the rem_pio2/ directory. For example, in rem_pio2/fpmath.h, there is
this comment: "Currently assumes Intel platform", followed by an
alternative that seems to assume either amd64 or i386.


Ah ok, thank you for considering.

I suppose it would be useful if we ran some build-time tests.
I think it would be better if the archs were not restricted and the 
package failed on the buildds.


Regards
Graham

Bug#799036: [Pkg-julia-devel] Bug#799036: Bug#799036: Bug#799036: openspecfun: please build everywhere

[Sébastien Villemot]

Le jeudi 17 septembre 2015 à 18:36 +0200, Graham Inggs a écrit :
> On 17/09/2015 17:38, Sébastien Villemot wrote:
> > I don't understand why this would be better. It seems to me that it
> > would just eat buildd resources, with no clear benefit… Or am I 
> > missing
> > something?
> 
> In my opinion, restricting architectures is hiding the problem.
> 
> In the case of openspecfun, it builds in seconds (even on ARM) so 
> wouldn't be much of a burden on the buildds.
> 
> If the build logs show a failure, it may attract the attention of 
> someone with an interest in that architecture and the problem could 
> get 
> fixed.

Quoting policy §5.6.8:

”Specifying a list of architectures or architecture wildcards other
than any is for the minority of cases where a program is not portable
or is not useful on some architectures. Where possible, the program
should be made portable instead.”

I consider that I'm using the architecture field in perfect compliance
with policy. The package is currently not portable, so I've restricted
the architecture field. Anyone is of course welcome to provide patches
to make it portable.

-- 
 .''`.Sébastien Villemot
: :' :Debian Developer
`. `' http://sebastien.villemot.name
  `-  GPG Key: 4096R/381A7594




signature.asc
Description: This is a digitally signed message part

Bug#799036: [Pkg-julia-devel] Bug#799036: openspecfun: please build everywhere

[Sébastien Villemot]

Le mardi 15 septembre 2015 à 08:50 +0200, Graham Inggs a écrit :
> Source: openspecfun
> Version: 0.4-1

> Please do not restrict openspecfun to i386 and amd64.
> It built on all architectures in Ubuntu [1].

It builds, but I am not sure it works. There is some low level stuff in
the rem_pio2/ directory. For example, in rem_pio2/fpmath.h, there is
this comment: "Currently assumes Intel platform", followed by an
alternative that seems to assume either amd64 or i386.

Viral, Eliott, what's your take on this? Does it make sense to compile
openspecfun on arches other than amd64 and i386?

-- 
 .''`.Sébastien Villemot
: :' :Debian Developer
`. `' http://sebastien.villemot.name
  `-  GPG Key: 4096R/381A7594




signature.asc
Description: This is a digitally signed message part

Bug#799036: [Pkg-julia-devel] Bug#799036: Bug#799036: openspecfun: please build everywhere

[Viral Shah]

Yes, we should restrict. Arm should be possible shortly. We really should
move the rem_pio2 to Julia so that this restriction can be lifted.

-viral
On 17 Sep 2015 2:21 am, "Sébastien Villemot"  wrote:

> Le mardi 15 septembre 2015 à 08:50 +0200, Graham Inggs a écrit :
> > Source: openspecfun
> > Version: 0.4-1
>
> > Please do not restrict openspecfun to i386 and amd64.
> > It built on all architectures in Ubuntu [1].
>
> It builds, but I am not sure it works. There is some low level stuff in
> the rem_pio2/ directory. For example, in rem_pio2/fpmath.h, there is
> this comment: "Currently assumes Intel platform", followed by an
> alternative that seems to assume either amd64 or i386.
>
> Viral, Eliott, what's your take on this? Does it make sense to compile
> openspecfun on arches other than amd64 and i386?
>
> --
>  .''`.Sébastien Villemot
> : :' :Debian Developer
> `. `' http://sebastien.villemot.name
>   `-  GPG Key: 4096R/381A7594
>
>
>
> ___
> Pkg-julia-devel mailing list
> pkg-julia-de...@lists.alioth.debian.org
> http://lists.alioth.debian.org/cgi-bin/mailman/listinfo/pkg-julia-devel
>

Bug#799036: openspecfun: please build everywhere

[Graham Inggs]

Source: openspecfun
Version: 0.4-1

Hi Maintainer

Please do not restrict openspecfun to i386 and amd64.
It built on all architectures in Ubuntu [1].

Regards
Graham


[1] https://launchpad.net/ubuntu/+source/openspecfun/0.4-1ubuntu1