[Gc] Re: gc6.8, etc.

Rainer Orth ro at TechFak.Uni-Bielefeld.DE
Thu Jul 13 13:04:01 PDT 2006


Hans,

> I decided to do a small amount of testing, and release the current gc6.X
> tree as 6.8.  It really only has a couple of additional ports and some
> small bug fixes relative to 6.7.

ok, thanks.  Checking 6.8, I notice that I've been mislead by the gc7.0 CVS
doc/README.changes which states

Since gc6.8:
[...]
 - Remove separate Solaris threads support.  Use the more generic Posix
   implementation.

It seems that those changes have only been done to the 7.0 version, despite
the heading.  I had hoped that those changes might be in 6.8 since they let
gctest on Solaris/x86 run successfully.  Unfortunately, those changes have
happened before 7.0 was imported into CVS, so I cannot extract the changes
from there either.  I'll see if I can extract them from the 7.0 sources
manually, but somewhat doubt that this is acceptable for GCC 4.2.0.

Maybe I'll just propose the Solaris/amd64 change so at least GCC bootstrap
works there, or even disable libgcj on that platform.

> I had hoped to package up a new gc7.0 version as well, but it still
> seems to have some problems with the dynamic library version on Windows.
> I still hope to do that, but it will be at least a day or two.  I will
> call that one 7.0alpha7 to distinguish it from the CVS versions.  I hope
> that it will be reasonably stable, MUCH more so than 7.0alpha5.  It has
> passed tests on various Linux(X86, IA64, X86-64, arm) and HP/UX
> platforms, MacOS/ppc, FreeBSD/X86, and Solaris/SPARC.
> 
> There will only be a 6.9 if I get one or more critical bug fixes.  I am
> no longer planning on including any new features or ports in 6.X.  Thus
> if you submit such patches, patches against 7.0, ideally cvs head, are
> greatly preferred.

Understood.  It makes sense to concentrate on a single release branch.  I
have already tried it on Tru64 UNIX, but ran into some issues I'll report
separately.  With two minor fixes, it works on V5.1B provided you don't use
GCC > 3.4 (cf. PR target/28307).  On V4.0F, the tests fail to link due to
their use of vsnprintf which doesn't exist on that release.

I'll repeat this for IRIX 5/6 once I get to it.

	Rainer

-----------------------------------------------------------------------------
Rainer Orth, Faculty of Technology, Bielefeld University


More information about the Gc mailing list