[Bug target/14344] [3.3.3 Regression] MinGW Bootstrap Failure

[stl at caltech dot edu]

------- Additional Comments From stl at caltech dot edu  2004-02-29 20:31 -------
[Gabriel Dos Reis]
> You reported that he produced a 3.3.3 MinGW patch set, then presumably
> he was able to bootstrap?

Yes, with Cygwin.

> I don't think that file did disappear in gcc-3.3-branch time frame.
> Rather it *appeared* in gcc-3.4.x with

Ok, that's what I guessed.

> Aaaaaaaarggh, I'm much more interested in *FSF* 3.3.3 than anything else.
> Please, oh yes, try that one.

Okay.

[Danny Smith]
> The config/i386/cygming.h in the mingw distribution of gcc-3.3.3 is a 
backport
> from gcc-3.4.  It does not exist in FSF sources of 3.3.3.  
> I am confused as to what sources you are using.  If you are using the FSF
> sources and have applied some of the mingw32-local patches but failed to add
> cygming.h than you would get the errors you see.

Above I have been using FSF sources from gcc.gnu.org patched with the MinGW 
patches from mingw.org.  This works for 3.3.1 but not for 3.3.3.

I will try using the components provided from mingw.org for 3.3.3 as well as
the pristine FSF sources.

> There is a sysroot-related bug that apppeared just before 3.3.3 was released,
> but that does not affect bootstap if everything is on same drive.  It only
> affects relocatabilty of toolset across disk volumes.

I work with C: only.

--

A couple of other issues:

When I take the FSF 3.3 sources and patch them with Danny Smith's 3.3 patch 
(sent
to a mailing list but never available from mingw.org), the bootstrap fails with
some message about stdio.h.  I remember fixing this somehow (certainly I did,
since I'm using a 3.3 I built myself right now), but I can't recall how.

I tried searching mailing list archives for this issue, since I recall it 
affecting
other people, but couldn't find anything. And this was only in May. Argh.

This wouldn't matter in the first place since I can build 3.3.1, except that 
3.3.1's
libstdc++.a is ten times the size of 3.3's!  That's approximately 9 MB versus 
900 KB.

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=14344