Message-Id: <199804091200.OAA45792@ieva06.lanet.lv>
From: "Andris Pavenis" <pavenis AT lanet DOT lv>
To: DJGPP Workers Mailing List <djgpp-workers AT delorie DOT com>,
        "John M. Aldrich" <fighteer AT cs DOT com>
Date: Thu, 9 Apr 1998 13:58:42 +0300
MIME-Version: 1.0
Content-type: text/plain; charset=US-ASCII
Content-transfer-encoding: 7BIT
Subject: Re: [Fwd: Re: What did I miss ?]
In-reply-to: <352C241A.32D3@cs.com>
Precedence: bulk

Date sent:      	Wed, 08 Apr 1998 21:27:54 -0400
From:           	"John M. Aldrich" <fighteer AT cs DOT com>
Subject:        	[Fwd: Re: What did I miss ?]

> Please take a look at the following log of gcc sent by a user on the
> DJGPP list.  Am I mistaken, or has the -dumpbase problem in gcc still
> not been corrected?
> 

Yes It is not corrected also in 2.8.1. 

gcc invokes cc1plus with command line argument '-dumpbase %b.cc' 
(fragment from specs, file cp/lang-specs.h). Therefore .o file still 
contains reference to source with extension .cc always even if source 
really had extension .cpp or something like. This problem is NOT DJGPP
specific.

I see some possible ways of fixing it:
       -  use '-dumpbase %i' instead of '-dumpbase %b.cc'
          then we'll get full path of source file. I'm not sure this
          will not break something.
       -  add extra specs code to get file extension.
          '%b' outputs input_basename, we should have code that outputs
          variable input_suffix (There is no such in gcc.c). Would 
          be nice I think.

Question: is it solved in egcs? I haven't tested it.