A funny error when compiling the gsl-1.2 candidate

A funny error when compiling the gsl-1.2 candidate

[Marek Kobera]

	Hello all!
I would not like to bother with a newbies' kind of question, but let me
ask
you about the cause of this error message, when doing make:
make[2]: *** Warning: File `diff/libgsldiff.la' has modification time in the fut
ure (2002-07-10 22:32:23.802911 > 2002-07-10 22:32:10.498715303)
make[2]: warning:  Clock skew detected.  Your build may be incomplete. 
We have got recently a new installation of SunOS 5.9:
SunOS 5.9 Generic sun4u sparc SUNW,Ultra-2
and gcc's version is 3.1.
	Thanks in advance,
		Marek Kobera

Re: A funny error when compiling the gsl-1.2 candidate

[Nicolas Bock]

On Wed, 10 Jul 2002, Marek Kobera wrote:

> 	Hello all! make[2]: *** Warning: File `diff/libgsldiff.la' has
> modification time in the fut ure (2002-07-10 22:32:23.802911 >
> 2002-07-10 22:32:10.498715303) make[2]: warning:  Clock skew detected.  
> Your build may be incomplete.

Are you mounting your build directory from an NFS server? I got this error 
message before too, and I believe, at least in my case, this had to do 
with the clocks set slightly differently on my machine vs. the NFS server.

	I hope this helps,

		nick

Re: A funny error when compiling the gsl-1.2 candidate

[Marek Kobera]

On Mon, 15 Jul 2002, Brian Gough wrote:

> Marek Kobera writes:
>  > 	Turning off the optimization solved the first three FAILs, but the
>  > last one still persists.
>  >
>  > >  > FAIL: test gsl_ran_ugaussian [4.2, 4.3) (4/100000 = 4e-05 observed
>  > >  > vs. 4.80585e-06 expected) // randist p.
>  > >  > P.S.: That's gcc 3.1 & SunOS 5.7.
> 
> I've tried this on on SunOS 5.8 with gcc 3.1 and the test passes ok.
> If you find the reason why it fails on your system let us know.
> 
> Brian
> 
	Please accept my apologies. I forgot that for the previous version
of GSL I had defined the GSL_RNG_TYPE to be 'taus'. That has caused the
error.
	Best regards,
		Marek

Re: A funny error when compiling the gsl-1.2 candidate

[Brian Gough]

Marek Kobera writes:
 > 	Turning off the optimization solved the first three FAILs, but the
 > last one still persists.
 >
 > >  > FAIL: test gsl_ran_ugaussian [4.2, 4.3) (4/100000 = 4e-05 observed
 > >  > vs. 4.80585e-06 expected) // randist p.
 > >  > P.S.: That's gcc 3.1 & SunOS 5.7.

I've tried this on on SunOS 5.8 with gcc 3.1 and the test passes ok.
If you find the reason why it fails on your system let us know.

Brian

Re: A funny error when compiling the gsl-1.2 candidate

[Marek Kobera]

	Turning off the optimization solved the first three FAILs, but the
last one still persists.
	Best regards,
		Marek Kobera

On Sun, 14 Jul 2002, Brian Gough wrote:

> Marek Kobera writes:
>  > FAIL: gsl_vector_long_double_div divides correctly (twice) // vector p.
>  > FAIL: gsl_matrix_long_double_div_elements divides correctly (twice) //
>  > FAIL: gsl_matrix_complex_long_double_div_elements divides correctly
>  > FAIL: test gsl_ran_ugaussian [4.2, 4.3) (4/100000 = 4e-05 observed
>  > vs. 4.80585e-06 expected) // randist p.
>  > 	
>  > 	Hope this helps somehow,
>  > 		Marek
>  > P.S.: That's gcc 3.1 & SunOS 5.7.
>  > 
> 
> Do the problems still occur if you compile without optimisation?
> If not, that would suggest a problem with the compiler.
> 
> regards
> Brian Gough
> 

Re: A funny error when compiling the gsl-1.2 candidate

[Brian Gough]

Marek Kobera writes:
 > FAIL: gsl_vector_long_double_div divides correctly (twice) // vector p.
 > FAIL: gsl_matrix_long_double_div_elements divides correctly (twice) //
 > FAIL: gsl_matrix_complex_long_double_div_elements divides correctly
 > FAIL: test gsl_ran_ugaussian [4.2, 4.3) (4/100000 = 4e-05 observed
 > vs. 4.80585e-06 expected) // randist p.
 > 	
 > 	Hope this helps somehow,
 > 		Marek
 > P.S.: That's gcc 3.1 & SunOS 5.7.
 > 

Do the problems still occur if you compile without optimisation?
If not, that would suggest a problem with the compiler.

regards
Brian Gough

Re: A funny error when compiling the gsl-1.2 candidate

[Marek Kobera]

	Dear Nicolas,
OK I have compiled the package on the NFS server - everything
(configure & make) all right,
but there were errors in make check, which I had partly encountered
before, as I
was trying to
compile it via NFS after several recompilations. I don't know if they have
been reported before (guess not), so here they are:
FAIL: gsl_vector_long_double_div divides correctly (twice) // vector p.
FAIL: gsl_matrix_long_double_div_elements divides correctly (twice) //
									matrix p.
FAIL: gsl_matrix_complex_long_double_div_elements divides correctly
(twice)//
									matrix p.
FAIL: test gsl_ran_ugaussian [4.2, 4.3) (4/100000 = 4e-05 observed
vs. 4.80585e-06 expected) // randist p.
	
	Hope this helps somehow,
		Marek
P.S.: That's gcc 3.1 & SunOS 5.7.

On Thu, 11 Jul 2002, Nicolas Bock wrote:

> On Thu, 11 Jul 2002, Marek Kobera wrote:
> 
> > 	Dear Nicolas and Trevor,
> > What should I be doing now? Should I 'make something', so that I build the
> > depcomp file or make check once again? 
> > I would really like to compile GSL on this machine and not on the NFS
> > server for a couple of reasons.
> 
> Marek,
> 
> quite frankly I don't know what this error means or why it is there, but 
> can you, just to check, compile gsl on a local drive and see if the tests 
> you mention fail? If they don't then we know that this is related to the 
> time stamp problem, and if they still fail then the problem is with the 
> tests and not with NFS.
> 
> 	Just an idea,
> 
> 		nick
> 
> 
> 

Re: A funny error when compiling the gsl-1.2 candidate

[Nicolas Bock]

On Thu, 11 Jul 2002, Marek Kobera wrote:

> 	Dear Nicolas and Trevor,
> What should I be doing now? Should I 'make something', so that I build the
> depcomp file or make check once again? 
> I would really like to compile GSL on this machine and not on the NFS
> server for a couple of reasons.

Marek,

quite frankly I don't know what this error means or why it is there, but 
can you, just to check, compile gsl on a local drive and see if the tests 
you mention fail? If they don't then we know that this is related to the 
time stamp problem, and if they still fail then the problem is with the 
tests and not with NFS.

	Just an idea,

		nick

Re: A funny error when compiling the gsl-1.2 candidate

[Marek Kobera]

> Marek Kobera wrote:
> > 	Hello all!
> > I would not like to bother with a newbies' kind of question, but let me
> > ask
> > you about the cause of this error message, when doing make:
> > make[2]: *** Warning: File `diff/libgsldiff.la' has modification time in the fut
> > ure (2002-07-10 22:32:23.802911 > 2002-07-10 22:32:10.498715303)
> > make[2]: warning:  Clock skew detected.  Your build may be incomplete. 
> > We have got recently a new installation of SunOS 5.9:
> > SunOS 5.9 Generic sun4u sparc SUNW,Ultra-2
> > and gcc's version is 3.1.
> > 	Thanks in advance,
> > 		Marek Kobera
> > 
> > 
> > 
> > 
> 
> That is a warning, not an error, I don't
> think it is something to worry about. You can
> maybe fix it by touching all the files:
> 
>    find . -type f -exec touch {} \;
> 
> 
	Dear Nicolas and Trevor,
thanks for your responses. I have realized that it is really the case with
the NFS server. However what I find interesting is that this warning shows
up by one file only, and that "Your build may be incomplete"-warning may
be true, since I have tried to 'make check' and received two specific
errors. Then I tried to build everything from the scratch again, used the
touch command form supposed by Trevor and made check once again. This, I
am afraid, does
not work. The error message is
Makefile.am: required file `./depcomp' not found
make: *** [Makefile.in] Error 1

which occurs just after
cd . && aclocal
cd . && automake --gnu --include-deps Makefile

What should I be doing now? Should I 'make something', so that I build the
depcomp file or make check once again? 
I would really like to compile GSL on this machine and not on the NFS
server for a couple of reasons.
	Hopefully you help me once again.
		Sincerely,
			Marek Kobera