From: khan AT xraylith DOT wisc DOT edu (Mumit Khan) Subject: Re: EGCS 1.0.2: weird include paths with gcc -b, specs, etc 22 Aug 1998 11:20:10 -0700 Message-ID: References: <6DA86DEBF68AD1119B1C00A0C94D230690987D AT hopper DOT logica DOT co DOT uk> Mime-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII To: "Fieldhouse, Dirk" Cc: gnu-win32 AT cygnus DOT com On Mon, 17 Aug 1998, Fieldhouse, Dirk wrote: > ref: Earnie Boyd's http://www.cygnus.com/ml/gnu-win32/1998-Apr/0028.html > > I'm trying to set up a dual cygwin32/mingw32 environment with the egcs 1.0.2 > cygwin32 binaries and being defeated by some strange behaviour with search > paths. > > The binary distribution of egcs-1.0.2 for cygwin32 seems to have some > hard-wired directories built into it, specifically > > ../../../../../include > and > ../../../../i386-cygwin32/include > Yeah, this is what I call the "Cygnus unlibsubdir" patch that is standard with Cygnus built distributions. Basically, unlibsubdir patch relocates the entire GCC package with only one env variable -- GCC_EXEC_PREFIX -- and everything is accessed relative to that. This change affects the language drivers, cpp and proto tools only. As with any choice of this type, there are pros and cons. I had built the first mingw32 release without this, and hence you needed all the other environment variables to run. I supplied a bat file to make life simple, and then ended up answering so many email messages that I got rather sick and tired of it. So, for egcs-1.0.2, and it will be so for egcs-1.1, I added this patch and haven't received a single message about not being able to run GCC due to incorrect variable setting. I believe Earnie was the only person to voice his opposition to this, but then again, he is more than a casual user, and fully capable of rebuilding from sources. I need to weigh my time against other factors, and unless someone provides a different mechanism, while keeping the "relocatability" feature, I'm afraid this feature (or bug, depending on who you are) is here to stay. What I *am* willing to do however is to build the drivers (gcc.exe, etc) without this patch and put it in a separate archive so you can use those instead. The only programs affected are the language drivers, cpp, protoize and unprotoize. BTW, the --exec-prefix and --prefix settings are different for cygwin32 (which use different values for these) and mingw32 (uses the same value for both), so what you see is consistent. Regards, Mumit - For help on using this list (especially unsubscribing), send a message to "gnu-win32-request AT cygnus DOT com" with one line of text: "help".