Re: regex.c not found, but it clearly exists

[Andrew STUBBS <andrew dot stubbs at st dot com>]

Dave Korn wrote:
>   This is basically it, but there are some problems.  The -mno-cygwin flag and
> the cross-compile mode of the cygming-targeted compiler are somewhat kludgey.
> The major purpose for which they are packaged is so that you can rebuild the
> cygwin dll itself from within the cygwin environment.  It's not really
> intended for general cross-development, and although it'll just about do,
> there is a problem: it doesn't correctly switch over /all/ the default search
> paths; if you compare the output of "gcc -v -E -xc - < /dev/null" and "gcc
> -mno-cygwin -v -E -xc - < /dev/null", you'll see that gcc -mno-cygwin still
> searches the cygwin compiler-specific include dir
>
> /usr/lib/gcc/i686-pc-cygwin/3.4.4/include
>
> when it should be looking in:
>
> /usr/lib/gcc/i686-pc-mingw32/3.4.4/include
>
>   I don't know - and couldn't say for sure without doing a thorough audit -
> how serious this could be, but from running a very quick diff over it, I can
> see that it might in particular affect C++.  (It could easily break passing
> std::strings across DLL boundaries, which cygwin has some custom mods to the
> c++ stl header files relating to).

We use it to build a complete cross-toolset - binutils, gcc, newlib, gdb 
and a few other bits and pieces. I've never had any trouble with search 
paths that I couldn't blame on the build system, although no doubt there 
are some pit-falls to be had.

Obviously there are a number of patches required to build all those 
utilities on windows. We also have a few patches to make the tools 
understand /cygdrive paths in a limited way - it won't build the target 
libraries without.

We used to target Cygwin, rather than Windows, but that meant that all 
our customers also had to use Cygwin, and there were version mismatch 
problems to overcome.

Andrew

P.S. Sorry about the multiple posts - I've received a notification from 
the mailserver saying that it's having trouble sending the message, but 
there doesn't seem to be any way to tell it not to bother.