Re: Problems building gcc-2.95.2

["Tim Prince" <tprince at computer dot org>]

Several interesting points were raised, which I am attempting to check by
building the standard gcc-2.95.3 myself right now.  After reading the
messages, I am trying ../configure
CC='/usr/bin/gcc -mwin32' --host=i686-pc-cygwin --target=i686-pc-cygwin --pr
efix=/usr/local/  in an attempt to avoid going into cross compiler mode.  I
find that my --prefix is being over-ridden to the extent of building in
/usr/local/i686-pc-cygwin, and that I will need to establish
/usr/local/i686-pc-cygwin/include in order for <windows.h> to be found.  It
seems that it is always necessary to put a copy of include in parallel with
the directory for which your compiler is being built; this is guaranteed to
be required when your target is named even slightly different from what your
cygwin installation uses.

In my previous attempts, the best I could do with configuring was to
bootstrap gcc.  If you get into the libf2c build and it fails because
[de]time_.c are consistent only with the cygwin-20 version of <windows.h>.
If I find obstacles in libf2c this time, I will simply copy a current
version over the one from gcc-2.95.2.  I had a somewhat acerbic exchange of
e-mails over the failure to update gcc-2.95.3 for current cygwin, which you
should be able to find in the gcc mail list archive.
 ----- Original Message -----
From: "Blythe.Stephen" <Stephen.Blythe@IGT.com>
To: "'Tim Prince'" <tprince@computer.org>; "'Cygwin Mailing List'"
<cygwin@cygwin.com>
Sent: Monday, February 19, 2001 9:52 AM
Subject: RE: Problems building gcc-2.95.2


>
> > 1st question:  if you are trying to build from the standard
> > source, rather
> > than the one from the cygwin mirror directories, why?
>
> Oh, I'm building from the standard source.  Is there a problem with doing
> this?  Has the cygwin source branched off from the standard source?

The cygwin source has several patches applied to make it consistent with the
current cygwin, as well as to make a number of options such as building
.dll's work.  My only reason for using the standard version is in order to
test gcc-2.95.3, which has some useful improvements.  I don't expect to get
a fully cygwin-functional build out of this, but I do expect a good basic
gcc and g77.  When I am done, I will update my page again at
members.aol.com/n8tm/cygwingccg77.htm
>
> > 2nd question:  are you sure that the target selected
> > automatically was not
> > i686-pc-cygwin32? It may be better to specify the target so that it is
> > identical to the host.  If your --host and --target do not
> > match exactly,
> > you are doing a cross build, which requires a copy of the
> > include files to
> > be present in the appropriate target build directory.
> >
> > 3rd question:  (depending on your answer to #2) did you
> > remember to copy
> > /usr/include to /usr/local/i686-gcc/i686-pc-cygwin32/include (if that
> > corresponds to the selected target) before running configure?
> >  That should
> > postpone problems finding <windows.h> until the libf2c build,
> > where you
> > really should be using the cygwin copy of the sources, even
> > if you choose
> > the standard gcc-2.95.2 or 2.95.3 .  The keepers of
> > gcc-2.95.3 have declined
> > the suggestion that libf2c should correspond with any version
> > of cygwin
> > other than -b20.
>
> Well, part of the makefile looks like this...
>
> ...
> host_alias = i686-pc-cygwin32
> host_cpu = i686
> host_vendor = pc
> host_os = cygwin32
> host_canonical = i686-pc-cygwin32
> target_alias = i686-pc-cygwin32
> target_cpu = i686
> target_vendor = pc
> target_os = cygwin32
> target_canonical = i686-pc-cygwin32
> build_alias = i686-pc-cygwin32
> build_cpu = i686
> build_vendor = pc
> build_os = cygwin32
> build_canonical = i686-pc-cygwin32
> ...
>
> ...so I assume that means my host and target are set ok, and yet the file
it
> fails on is prefix.c so who knows?
>
> Can you explain, or point me to something which explains, you comments on
> libf2c?
>
> Cheers,
>
> Stephen
>


--
Want to unsubscribe from this list?
Check out: http://cygwin.com/ml/#unsubscribe-simple