Re: libstdc++/4146: input, num_get, regression from 2.95

[bkoz at gcc dot gnu dot org]

The following reply was made to PR libstdc++/4146; it has been noted by GNATS.

From: bkoz@gcc.gnu.org
To: bkoz@gcc.gnu.org, bkoz@gnu.org, gcc-bugs@gcc.gnu.org,
  gcc-gnats@gcc.gnu.org, gcc-prs@gcc.gnu.org, wall@chop.swmed.edu
Cc:  
Subject: Re: libstdc++/4146: input, num_get, regression from 2.95
Date: 1 Dec 2001 02:29:28 -0000

 Synopsis: input, num_get, regression from 2.95
 
 State-Changed-From-To: analyzed->feedback
 State-Changed-By: bkoz
 State-Changed-When: Fri Nov 30 18:29:28 2001
 State-Changed-Why:
     Hey. I just examined this with current gcc-3.1 head. Sorry for the delay, by the way.
     
     You're right, it was very simple:
     
     in read_pdb_file, the ifstream is opened, read, closed, then re-opened. This is all fine, but the file is reopened without clearing error bits, thus failing to deal with the resulting data.
     
     On line 197 of my copy of these sources:
     
      pdb_file.open(pdb_file_name, std::ios::in);
       pdb_file.clear();   // XXX bkoz DR 22 need to clear eofbit.
       while (pdb_file >> dummy) 
      
     Adding the clear solves the problem....
     
     Anyway. Try this: it should solve your problem.
     
     I noticed that the output file is formatted a bit differently between gcc-2.96 and gcc-3.x:
     
     2.96:
     ATOM      2  CA  ASN   584    22.622  -18.771 28.061  1.00  30.76 
     
     gcc-3.x:
     ATOM      1  N   ASN   584      23.122 -19.077  26.733  1.00 31.16
     
     
     Notice the different spacing right before the floating poin type. I believe this is correct:
     
     align (right) << "    " // 4 spaces
     setwidth (8) << output 6-digit numeric type.
     
     Looks right to me.
     
     Here's hoping this solves your problem.
 
 http://gcc.gnu.org/cgi-bin/gnatsweb.pl?cmd=view%20audit-trail&pr=4146&database=gcc