whizard is hosted by Hepforge, IPPP Durham

Opened 10 years ago

Closed 10 years ago

#654 closed defect (fixed)

Fix seg faults in tests (both with NAG and gfortran)

Reported by: Juergen Reuter Owned by: kilian
Priority: P0 Milestone: v2.2.3
Component: core Version: 2.2.2
Severity: critical Keywords:
Cc:

Description (last modified by Juergen Reuter)

There is a problem with the NAG compiler at the moment: 

/bin/bash ../../libtool --tag=FC   --mode=compile nagfor -I../misc -I../../vamp/src -I../shower -I../pdf_builtin -I../muli -I../../circe1/src -I../../circe2/src -I../lhapdf -I../fastjet   -C=all -nan -gline -w -f2003 -maxcontin=512 -c -o fks_regions.lo  fks_regions.f90
libtool: compile:  nagfor -I../misc -I../../vamp/src -I../shower -I../pdf_builtin -I../muli -I../../circe1/src -I../../circe2/src -I../lhapdf -I../fastjet -C=all -nan -gline -w -f2003 -maxcontin=512 -c fks_regions.f90  -PIC -o .libs/fks_regions.o
NAG Fortran Compiler Release 5.3.1(934)
Error: fks_regions.f90, line 177: LIST is neither a POINTER nor a TARGET
Error: fks_regions.f90, line 225: LIST is neither a POINTER nor a TARGET
Error: fks_regions.f90, line 246: LIST is neither a POINTER nor a TARGET
Error: fks_regions.f90, line 487: Missing comma in format specification
Error: fks_regions.f90, line 489: Missing comma in format specification
Error: fks_regions.f90, line 1039: Missing comma in format specification
Error: fks_regions.f90, line 1053: Missing comma in format specification
[NAG Fortran Compiler error termination, 7 errors]
make[3]: *** [fks_regions.lo] Error 1

After r6006, WHIZARD does compile with NAG again, but 9 tests are failing now: 

phs_forests.run
phs_wood.run
cuts.run
smtest_1.run
smtest_2.run
smtest_6.run
smtest_7.run
mlm_matching_isr.run
qedtest_4.run

Change History (8)

comment:1 Changed 10 years ago by Juergen Reuter

Description: modified (diff)
Severity: normal → critical
Summary: trunk does not compile with NAG → Fix tests with NAG compiler

comment:2 Changed 10 years ago by Juergen Reuter

Summary: Fix tests with NAG compiler → Fix seg faults in tests (both with NAG and gfortran)

comment:3 Changed 10 years ago by Juergen Reuter

The error in phs_forests is definitely something broken: 

Runtime Error: phs_trees.f90, line 1210: Subscript 1 of TREE%BRANCH (value 0) is out of range (1:31)
Program terminated by fatal error
phs_trees.f90, line 1210: Error occurred in PHS_TREES:PHS_TREE_GET_ANGLES:GET_ANGLES_X
phs_trees.f90, line 1116: Called by PHS_TREES:PHS_TREE_GET_ANGLES
phs_trees.f90, line 995: Called by PHS_TREES:PHS_TREE_COMPUTE_X_FROM_MOMENTA
phs_forests.f90, line 1317: Called by PHS_FORESTS:PHS_FOREST_EVALUATE_OTHER_CHANNELS
phs_forests.f90, line 1477: Called by PHS_FORESTS:PHS_FOREST_1
unit_tests.f90, line 176: Called by UNIT_TESTS:TEST
phs_forests.f90, line 1358: Called by PHS_FORESTS:PHS_FOREST_TEST
whizard.f90, line 563: Called by WHIZARD:WHIZARD_CHECK
main.f90, line 355: Called by MAIN
Aborted

comment:4 Changed 10 years ago by kilian

Owner: changed from kilian, cweiss to kilian
Status: new → assigned

WK: entering debugging mode ...

NAG runtime crash reproduced

comment:5 Changed 10 years ago by kilian

Fixed trivial typo ... rerunning testsuite

comment:6 Changed 10 years ago by kilian

Resolution: → fixed
Status: assigned → closed

r6010: works with NAG. The gfortran tests were already ok.

... leaving debugging mode. Closing.

comment:7 Changed 10 years ago by Juergen Reuter

Resolution: fixed
Status: closed → reopened

Not yet all tests working, at least the LHAPDF, lhapdf and sf_lhapdf tests are still failing. Back to work. Sorry.

comment:8 Changed 10 years ago by Juergen Reuter

Resolution: → fixed
Status: reopened → closed

This was broken code from a malformed merge of the NLO branch. Fixed in r6013. All seems fine again. Closing.

Note: See TracTickets for help on using tickets.