GCC 3.4.2 Status (2004-08-23)

[Mark Mitchell <mark at codesourcery dot com>]

--
Mark Mitchell
CodeSourcery, LLC
(916) 791-8304
mark@codesourcery.com

This message addresses only the upcoming GCC 3.4.2 release, not the
GCC 3.5 release.  I will have more in that over the next few days, now
that the final project submissions should be in.

The GCC 3.4.2 release date remains targeted at September 1st.  I plan
to spin pre-release tarballs on Saturday, August 28th, and make the
final release on the 1st, assuming no catastrophic problems.

There are, however, still 56 (!) wrong-code, rejects-valid, or
ice-on-valid-code regressions targeted GCC 3.4.2.  My plan is to
basically move most of these back, as people don't seem very concerned
about them, and very few (none?) are regressions from previous 3.4.x
releases.  

However, I'm disturbed by the lack of attention these bugs are
getting; many are platform-independent, and many others deal with
popular features like SSE2 or debugging output, and most seem to occur
in 3.5 as well.

Our bugmasters are working very hard to make it easy for us developers
to have nice test-cases to work with, so I think that with relatively
little effort, we could be closing a lot more PRs than we are.

I'm also disappointed by the fact that there are a few relative
important regressions that, despite pleas for help targeted at
particular developers, have not resulted in the aforementioned
help. :-(

Here they are.  If your name is on one of these, would you please at
least indicate whether or not you are going to be able to look at this
in short order?

PR 16590

  Problem with GCSE, patch submitted.

  Jeff Law, August 17th

PR 15461

  Problem with the C++/NRV implementation.  

  Jason Merrill, June 22nd

PR 16851

  Problem with stabilize_init

  Jason Merill, August 17th

PR 15289

  Problem with reload, possibly related to one of Jan's patches,
  although that patch may or may not really be a problem.

  Jan Hubicka, August 19th.

There's also PR 14492, a crash in DWARF2, which I asked Jason about on
May 31, 2004.  This is somewhat less critical, in that not using -g is
a workaround, but it's not a *good* workaround.

I also recently assigned two PRs to Roger Sayle for middle-end
issues.  Roger, would you please indicate whether or not you'll be
able to work on those?  (These are 16693 and 17078.)

There seem to be a lot of SH regressions.  I'm not too concerned,
since this isn't a release-critical platform, but I'd like to see
Joern and Alexandre spending some time to try to clean some of these
failures up.

Thanks,

--
Mark Mitchell
CodeSourcery, LLC
(916) 791-8304
mark@codesourcery.com