Re: Testing gcc 4.2.3

[Brian Dessent <brian at dessent dot net>]

Reini Urban wrote:

> About the shared libgcc:
> Since this problem will not go away until we decide to keep the old and slow
> WINAPI compatible SJLJ (hopefully not), we can just add this libgcc as
> Base package.
> Similar to the msvcrt.dll in the windows world.

That would be fine if it were even possible to build a shared libgcc. 
FSF gcc won't even do this, nor will it build shared libstdc++, etc. 
MinGW's gcc 4.2 has a humongous local patch that backports a bunch of
stuff from 4.3 as well as using an ugly kludge to build shared libgcc.

I can't speak for Dave or whoever else will spend time maintaining the
Cygwin gcc packages, but I can say that if the past is any indication,
maintaining a huge local out-of-tree patch is a giant pain and causes a)
releases/updates to be slow or nonexistant and b) maintainers to not
want to continune doing their thing.  It's for these reasons that I
think we really need to wait until these issues are fixed upstream so
that we do not repeat what occured with the 3.x series, wherein we have
currently have a set of local patches that are 548KB with a diffstat of:
154 files changed, 4884 insertions(+), 1669 deletions(-).  (And those
numbers don't include all the additions for adding the out-of-tree
languages D and Pascal, so the actual changes are even greater.)

Brian

--
Unsubscribe info:      http://cygwin.com/ml/#unsubscribe-simple
Problem reports:       http://cygwin.com/problems.html
Documentation:         http://cygwin.com/docs.html
FAQ:                   http://cygwin.com/faq/