Bug#846353: FTBFS: ERROR: Test "ruby2.3" failed. Exiting.

[Gilles Filippini]

Gilles Filippini a écrit le 03/12/2016 à 01:35 :
> Gilles Filippini a écrit le 02/12/2016 à 23:26 :
>> Sebastiaan Couwenberg a écrit le 02/12/2016 à 15:34 :
>>> On 12/02/2016 09:55 AM, Sebastiaan Couwenberg wrote:
>>>> On 12/01/2016 07:48 PM, Sebastiaan Couwenberg wrote:
>>>>> On Wed, 30 Nov 2016 15:10:03 +0000 Iain Lane wrote:
>>>>>> This package FTBFS on 32 bit arches (works on amd64 and other 64 bit
>>>>>> arches), maybe due to the new hdf5
>>>>>
>>>>> Quite likely, since hdf-eos5 hasn't been rebuilt for the hdf5 transition
>>>>> yet. I suspect this issue to fix itself once that's done.
>>>>
>>>> It did not, but did only fail on 32-bit architectures.
>>>
>>> Gilles, these failures seem to be caused by the hid_t type change from
>>> 32-bit to a 64-bit value. I'm not sure how to fix it for ruby-hdfeos5,
>>> can you maybe provide a patch?
>>
>> Damn, I tested the build on amd64 only :/
>> I've tried fixing the 'incompatible pointer type' warning, but it
>> deosn't change anything wrt the test suite.
>> Is there a way to ensure first that hdf-eos5 works fine on 32 bit archs?
> 
> CC-ing Alastair, maintainer of hdf-eos5.
> 
> AIUI the problem lies in hdf-eos5. I've ran a very simple C test made
> after the first lines of test/test_gd.rb (source attached), both on
> amd64 and i386. It succeeds on amd64 and fails on i386 at the
> HE5_GDdetach step with:
> 
> $ ./myhe5-32
> bufsize = 19
> ColumnAmountAerosol
> HDF5-DIAG: Error detected in HDF5 (1.10.0-patch1) thread 4142823168:
>   #000: GDapi.c line 1140 in HE5_GDchkgdid(): Cannot get the Grid index
> 
>     major: Invalid arguments to routine
>     minor: Out of range
> HDF5-DIAG: Error detected in HDF5 (1.10.0-patch1) thread 4142823168:
>   #000: EHapi.c line 2206 in HE5_EHchkfid(): Invalid file ID: 0. ID
> should range from 67108864 to  67109064 .
> 
>     major: Invalid arguments to routine
>     minor: Bad value
> HDF5-DIAG: Error detected in HDF5 (1.10.0-patch1) thread 4142823168:
>   #000: GDapi.c line 1151 in HE5_GDchkgdid(): Checking for file ID failed.
> 
>     major: Invalid arguments to routine
>     minor: Out of range
> HDF5-DIAG: Error detected in HDF5 (1.10.0-patch1) thread 4142823168:
>   #000: GDapi.c line 12623 in HE5_GDdetach(): Checking for grid ID failed.
> 
>     major: Invalid arguments to routine
>     minor: Out of range

The problem seems to be the type conversion functions in EHapi.c. For
instance:

long
HE5_EHhid2long(hid_t invalue)
{
  long  outvalue = FAIL;

  if (sizeof(hid_t) == sizeof(int))
        outvalue = HE5_EHint2long(invalue);
  else if (sizeof(hid_t) == sizeof(long))
        outvalue = invalue;

  return(outvalue);
}

On 32 bit archs:
  sizeof(hid_t) != sizeof(int)
  sizeof(hid_t) != sizeof(long)
  sizeof(hid_t) == sizeof(long long)

The conversion functions should be adapted accordingly.

Thanks,

_g.

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 480 bytes
Desc: OpenPGP digital signature
URL: <http://lists.alioth.debian.org/pipermail/pkg-grass-devel/attachments/20161203/e9239182/attachment.sig>