From: softbrek AT POOL DOT Informatik DOT RWTH-Aachen DOT DE
Subject: You can't read iostream.inf with info.exe? read this!
To: djgpp AT sun DOT soe DOT clarkson DOT edu (djgpp)
Date: Wed, 23 Nov 1994 11:35:32 +0100 (MET)
Cc: dj AT ctron DOT com (DJ Delorie)

Hello, All (and esp. DJ!)

This is a bug-report for the file "docs/libgpp/iostream.inf" in the
current lgp260dc.zip, and the proper fix is given. (DJ: this ought
to be fixed in the next maint. rel., I think)

Symptoms
This monday, Jerry van Dijk (jerry AT jvdsys DOT textlitho DOT nl), posted
to this list, having a problem with iostream.inf from lgp260dc.zip, 
the documentation for FSF's lib-c++ : info.exe, the on-line
documentation browser, refused to display information on the
iostream-library, saying it "couldn't find the node 'TOP'".

Diagnostics:
============

This is really a bug in the distributed file '.../iostream.inf', as
contained in lgp260dc.zip (at least on our local SIMTEL-mirror): that
file has somehow been translated to DOS-style line-endings.

More Diagnostics (in case that didn't ring any bell in your mind :-):
=====================================================================

The distributed version of info chokes on such a file, iff it's a
split info-file (iostream.inf, iostream.i1, iostream.i2 all together
are the proper iostream.info, and theyre crosslinked via *absolute
file positions*, for faster cross-referencing.  These absolute
file-positions are *broken* by the CR's inserted at every LF. That
causes info.exe to get lost in the file's structures and barf at 
the poor user, since it can't find even the startup node, 'Top'.

The cure (for now):
===================

cd to .../docs/libgpp and type the following words of wisdom:

	dtou iostream.inf  (Return)

That's it.

The cure (for ever, but a bit harder to do):
============================================

Make info.exe strip these CR's after reading, and circumvent its
file-size checking routine, so it knows that the difference between
reported file-size and number of characters read (after having
stripped the CR's) is legal. (This was working once, but this
particular part of the port seems to have got lost on the way
from the original porter (ME), to DJ's first 'official' release.
(I still have the original diff's, so we could easily put it back
in the sources.)

This would stop this possible cause of error once and for ever,
reducing the amount of possible user confusion a bit, I think

Hoping this therapy works:

Hans-Bernhard Broeker (softbrek AT POOL DOT Informatik DOT RWTH-Aachen DOT DE)
                      (Broeker AT axpmgr DOT physik DOT rwth-aachen DOT de)