Processed: Re: Bug#1017862 scipy: FTBFS on 32 bit arches: TestMLS: Invalid call to pythranized function

[Debian Bug Tracking System]

Processing control commands:

> notfixed 1017862 1.8.1-8
Bug #1017862 [src:scipy] scipy: FTBFS on 32 bit arches: TestMLS: Invalid call 
to pythranized function
No longer marked as fixed in versions scipy/1.8.1-8.
> severity 1017862 important
Bug #1017862 [src:scipy] scipy: FTBFS on 32 bit arches: TestMLS: Invalid call 
to pythranized function
Severity set to 'important' from 'serious'

-- 
1017862: https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=1017862
Debian Bug Tracking System
Contact ow...@bugs.debian.org with problems

Bug#1017862: scipy: FTBFS on 32 bit arches: TestMLS: Invalid call to pythranized function

[Drew Parsons]

Control: notfixed 1017862 1.8.1-8
Control: severity 1017862 important

There are still two remaining pythran failures that need to be fixed in 
PR16646.


1.8.1-8 just ignores the problem, skips the failing tests. Hence I don't 
consider the bug fixed yet.


Ignoring the problem means the severity can be lowered from Serious to 
Important. It's not Release Critical now.


I want to wait for the full fix in PR16646 before closing the bug.

Drew

Bug#1017862: scipy: FTBFS on 32 bit arches: TestMLS: Invalid call to pythranized function

[Drew Parsons]

Source: scipy
Version: 1.8.1-8
Severity: serious
Justification: FTBFS
Control: forwarded -1 https://github.com/scipy/scipy/pull/16646

scipy FTBFS on 32 bits arches after pythran support was recently
activated. The error looks like:

___ TestMLS.test_mls_inputs 
/<>/.pybuild/cpython3_3.10_scipy/build/scipy/signal/tests/test_max_len_seq.py:21:
 in test_mls_inputs
assert_array_equal(max_len_seq(10, length=0)[0], [])
self   = 
/<>/.pybuild/cpython3_3.10_scipy/build/scipy/signal/_max_len_seq.py:136:
 in max_len_seq
state = _max_len_seq_inner(taps, state, nbits, length, seq)
E   TypeError: Invalid call to pythranized function 
`_max_len_seq_inner(intc[:], int8[:], int, int, int8[:])'
E   Candidates are:
E   
E   - _max_len_seq_inner(int[:], int8[:], int, int, int8[:])
E   - _max_len_seq_inner(int64[:], int8[:], int, int, int8[:])
length = 0
n_max  = 1023
nbits  = 10
seq= array([], dtype=int8)
state  = array([1, 1, 1, 1, 1, 1, 1, 1, 1, 1], dtype=int8)
taps   = array([7], dtype=int32)


Likewise with TestMLS.test_mls_output.

1.8.1-8 introduced debian patch pythran_type_PR16646.patch which
applied scipy upstream PR16646 to fix such pythran issues. The patch
fixed the problems reported at
https://github.com/serge-sans-paille/pythran/issues/2002
but two tests in TestMLS remain to be fixed.

This is the main problem keeping scipy 1.8 from migrating to testing,
hence marking bug severity Serioes.