Re: preprocessor/2948

[Sean McNeil <sean@mcneil.com>]

The following reply was made to PR preprocessor/2948; it has been noted by GNATS.

From: Sean McNeil <sean@mcneil.com>
To: Neil Booth <neil@daikokuya.demon.co.uk>
Cc: nobody@gcc.gnu.org, gcc-gnats@gcc.gnu.org
Subject: Re: preprocessor/2948
Date: Tue, 29 May 2001 09:40:39 -0700

 The pragmas are indeed getting passed through by the cpp0.  So why would
 they compile differently in the two cases?  This is very strange.  It is
 related to that #pragma, though.  Perhaps the cpp is told to parse things
 differently?
 
 Sean
 
 Neil Booth wrote:
 
 > Sean McNeil wrote:-
 >
 > > This has puzzled me for a while and I've tried several additional
 > > things as we've been discussing this.  I noticed that the cc1plus
 > > program installs handlers for various pragmas and discovered a
 > > #pragma implementation in PlotFile.cpp.  When I remove this pragma,
 > > compilation is as I expected.
 >
 > Great, thanks for persisting.
 >
 > > HOWEVER....  there is no template code that exists within the ttt.ii
 > > file (after preprocessor) and adding a #pragma implementation to
 > > ttt.cc (copy of ttt.ii) doesn't reproduce the problem either.  So I
 > > am still at a loss as to why
 >
 > Hmmm.  This doesn't sound good.  Are you saying that the #pragma does
 > not exist in the -E preprocessed output?  If so, that's a bug.  Most
 > pragmas are not handled or understood by CPP, so it must pass them
 > through in the preprocessed output unchanged.
 >
 > > 1) there is a #pragma implementation in this file (Something I do
 > > not expect anyone on > the gcc team to answer :).
 >
 > I have no idea.  I'm not really sure what the pragma does, either,
 > though I know it's to do with template instantiation.
 >
 > > 2) why code is generated for the pragma even though
 > > -fexternal-templates was not defined.
 >
 > Again, this would require me to understand exactly what it does.
 >
 > > I suppose what it boils down to is it's not a cpp problem, but a
 > > cc1plus problem.  Should this be reassigned?
 >
 > Probably, provided that CPP is passing through the pragma.  I'll
 > do that when we get just a little closer.
 >
 > Does this code work, both ways, with 2.95.2 or 2.95.3?  If so, it's a
 > regression, and I'll mark it high priority and it should get done for
 > 3.0.
 >
 > Neil.