Re: Chained modules in a testcase

Re: Chained modules in a testcase

[Harald Anlauf]

Hi Paul,

I got a list of potential candidates by

% grep 'dg-additional-sources .*f90 ' gcc/testsuite/gfortran.dg/ -r

and ran the first one:

% make check-fortran RUNTESTFLAGS='dg.exp=altreturn_9*.*'

On first sight this looked fine.

Cheers,
Harald

Chained modules in a testcase

[Paul Richard Thomas]

[-- Attachment #1: Type: text/plain, Size: 850 bytes --]

Hi,

Andrew Benson found a rather horrible bug in my finalization patch that
involved two separate module files and another containing both a module and
the main program. They must be compiled separately for the bug to appear.
This bug arises in finalization because some derived type function results
wind up with a derived type that has no proc_name for its namespace, which
causes gfc_get_derived_vtab to segfault.

file 1: iso.f90 which contains a module only.
file 2: dt.90 which uses module iso within one of the module contained
functions
file 3: ov.f90 which contains another module using iso and dt, again within
a module constrained function. It also contains the main program.

How do I deal with this in the testsuite? I have tried all manner of
combinations of dg-compile-aux-modules and dg-additional-sources to no
avail.

Cheers

Paul