[patch, libgfortran] [4.3/4.4/4.5 Regression] reading array of structures from namelist fails

[Jerry DeLisle]

Hi all,

This patch fixes this regression by adding an additional flag to identify when 
array qualifiers have been seen for a derived type component.  This flag is then 
used to identify when to reset the current namelist pointer to the correct 
namelist pointer (i.e. setting nl to first_nl).

The patch also cleans up the code by consolidating these flag tests in one place 
just before the call to read the object.  I attempted to clean this up so that 
it is most readable.  The logic is somewhat subtle to say the least.

As anyone figured out the half-life decay time for namelist bugs? :)

Regression tested on x86-64.  New test case provided.

OK for trunk and then back port to 4.4 and 4.3?

Regards,

Jerry

2010-02-02  Jerry DeLisle  <jvdelisle@gcc.gnu.org>

	PR libfortran/42901
	* io/list_read.c (nml_get_obj_data): Add new qualifier flag, clean up
	code, and adjust logic to set namelist info pointer correctly for array
	qualifiers of derived type components.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: submit.diff
Type: text/x-patch
Size: 2425 bytes
Desc: not available
URL: <http://gcc.gnu.org/pipermail/gcc-patches/attachments/20100203/b21bba32/attachment.bin>
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: namelist_60.f90
URL: <http://gcc.gnu.org/pipermail/gcc-patches/attachments/20100203/b21bba32/attachment.f90>