gcc-4.0.2

[gerard@…]

Added

[gerard@…]

Downgrading book to gcc-2.95.3 and change official url of this package to ftp.gnu.org/pub/gnu/gcc/

[gerard@…]

Done.

[markh@…]

It's here, it's ready and it's in conglomoration for your testing pleasure ;-)

[highos@…]

I would like to discuss the possibility of removal of '--with-slibdir=/lib' from LFS-BOOK CVS. I can't seem to find where this was added to CVS, nor the ChangeLog entry in the book. However, I do remenber all the various dicussions that took place on the mailing lists. Teemu Tervo and myself have noticed that no binaries inside /bin or /sbin actually link to libgcc_so.so and such only bin's in /usr/bin or /usr/sbin do. So, perhaps this is stuff left over from previous versions of gcc3. We are both running LFS-CVS ranging from a month to two weeks old (thou pretty much the same, considering nothing has changed much since I did LFS CVS) Doing a quick 'ldd * | grep libgcc_s' inside /bin and /sbin should be enough, to give us an idea if this is needed anymore in LFS CVS.

[highos@…]

"From: Greg Schafer <gschafer@…>

​http://gcc.gnu.org/ml/gcc/2002-02/msg01173.html

So in essence, when we moved to binutils-2.12.x, we gained PT_GNU_EH_FRAME and from that point onwards C binaries don't link against libgcc_s.so but C++ binaries still do."

Either way, there's no real harm in leaving it as-is like so, and you never know if someone perhaps needs to put a C++ app in /bin&/sbin so leaving it "as-is" should be enough.

Yay for curiosity ;)

[markh@…]

NOTE: gcc-3.1.1 isn't out yet, this is just a pointer about it.

I'm not sure we should bother moving the book to gcc-3.1.1. Reasoning? An e-mail to the gcc mailing list from Mark Mitchell (the release co-ordinator). Here's the relevant excerpt:

" 2. There will be no 3.1.2 release. Instead, there will be a GCC 3.2

release which is GCC 3.1.1 with those C++ ABI fixes present on the mainline. This release will take place shortly after GCC 3.1.1. (It will happen approximately as quickly as I can create branches and Jason can commit the relevant fixes.)"

So basically, we'll do the gcc-3.1.1 testing, put it in CVS and then, a week or so later (by the sounds of it) have to start again with gcc-3.2.

I think we should just await gcc-3.2.

The full email is available at: ​http://gcc.gnu.org/ml/gcc/2002-07/msg00596.html if you're interested.

[highos@…]

Version increment. I haven't bothered to upload a copy. Like Marh mentioned, we may as well just wait untill 3.2 gets released at this point.

[highos@…]

* Bug 383 has been marked as a duplicate of this bug. *

[highos@…]

Version increment. I was thinking of uploading a copy and then someone on #LFS mentioned that I was going to be evil and just put a copy of core&g++, so I think i'll just let Gerard/Mark/Timothy do the honour.. ;)

[gerard@…]

Can't work on it right now. Will try later, someboyd else feel free to take over.

[markh@…]

A copy has been put in /home/ftp/lfs-packages/conglomoration

When we move to gcc-3.2 suggest we add:

--enable-cxa_atexit see ​http://www.zipworld.com.au/~gschafer/lfs-tweaks.html and ​http://gcc.gnu.org/gcc-3.2/c++-abi.html

for reasoning.

We also need to sort out which glibc-2.2.5 patches are needed for use with the gcc-3.2/glibc-2.2.5 combination. See bug #386 for more details and also Greg's patch page listed above.

[highos@…]

Version increment (3.2.1)

[highos@…]

Version increment (3.2.2)

[highos@…]

Version increment (3.2.3)

[highos@…]

Version increment (3.3)

[highos@…]

Version increment (3.3.1)

[jeremy@…]

Book has been updated to GCC 3.3.1 - finishing touches are in process.

[greg@…]

Version increment (3.3.2)

[greg@…]

The new test results URL is here:

​http://gcc.gnu.org/ml/gcc-testresults/2003-10/msg00803.html

• 1 XPASS (unexpected pass) for g++
• 1 FAIL for gcc
• 26 XPASS's for libstdc++

[tushar@…]

Patches uploaded.

[greg@…]

Upgraded book to 3.3.2. Closing.

[winkie@…]

Version increment (3.3.3) w00t!

[pinskia@…]

Note that 3.4.0 should be out in the next two or three weeks, this is only an estimate as the release manager only knows for sure.

[ryan.oliver@…]

3.4 may be here soon, but as it stands presently it cannot compile glibc. (This will probably be fixed shortly).

[Matthew Burgess]

Jim Gifford mentioned that there's a problem caused by combining binutils-2.15.91.0.2 and gcc-3.4.1. It's a known bug in gcc (​http://gcc.gnu.org/bugzilla/show_bug.cgi?id=16625) and Jim's testing a patch for gcc at the moment. Alternatively we could revert back to binutils-2.15.90.0.3 (the last version that doesn't generate comdat and linkonce sections in the same line according to the changelog) or to binutils-2.15 (FSF).

[jim@…]

It looks to be a two patch scenario, one for binutils and one for gcc. Will have patches posted as soon as testing is completed.

[jim@…]

Testing patch recommended by GCC maintainers.

​http://tinyurl.com/686pb

I have asked the person who was having Mozilla build issues to try this patch also. I'm currently testing it on the kdegraphics package.

[jim@…]

Patch fixed the kdegraphics issue I had. Still no reports from people building mozilla as of yet. Will start a build myself in the am.

[jim@…]

Just received a report for a Mozilla compile. This issue is fixed with the patch that I just placed into the patches SVN.

Please advise if we want to put this patch into the BOOK, or wait until 9-1-2004 for GCC 3.4.2?

[Matthew Burgess]

The patch is now in the book, and LFS-6.0 will ship with gcc-3.4.1. Marking as fixed.

[jim@…]

Notes - Actual GCC Bugzilla # on the linkone/comdat issue is 16276.

This issue has been postponed until the 3.4.3 release. Additional Comment #22 From Mark Mitchell 2004-08-23 21:00 [reply] Postponed until GCC 3.4.3.

​http://gcc.gnu.org/bugzilla/show_bug.cgi?id=16276

The patch that I created is based on the information in this ticket.

[Matthew Burgess]

Version increment (3.4.2). Need to check the linkonce patch is still required and that the util-linux 'sed' is still needed. Grub's testsuite also showed a gcc bug. I doubt any of these bugs have been fixed in this patch release, but it'd be nice to confirm.

[jim@…]

All GCC patches have been converted to the current patch format and rediffed.

[jeremy@…]

Testsuite results at:

​http://gcc.gnu.org/ml/gcc-testresults/2004-09/msg00261.html

Will be going into unstable tonite.

[jeremy@…]

util-linux sed still needed, even with util-linux 2.12c.

[jeremy@…]

Upgraded in Unstable at rev 4142. Testing is in version freeze, so this won't be applied there. Closing.

[Matthew Burgess]

Version increment (3.4.3)

[jim@…]

The linkonce and comdat stuff has been delayed again. ​http://gcc.gnu.org/bugzilla/show_bug.cgi?id=16276

Additional Comment #28 From Mark Mitchell 2004-10-31 01:03 [reply] Postponed until GCC 3.4.4.

[jim@…]

Current linkonce patch doesn't work, will have the new one in patch in 2 hours, after testing is completed.

[jeremy@…]

Upgraded in unstable.

[Matthew Burgess]

Version increment (4.0.0). Please use the gcc-4 branch for upgrading this!

[Matthew Burgess]

* Bug 1062 has been marked as a duplicate of this bug. *

[Matthew Burgess]

Version increment (4.0.1)

[Matthew Burgess]

gcc4 branch merged to trunk in r6819, therefore marking as fixed...finally :)

[Matthew Burgess]

Version increment (4.0.2)

[Matthew Burgess]

gcc-4.0.2 is broken on m68k (does cross-lfs do m68k yet?) and has some libstdc++ issues on Solaris (obviously doesn't affect us). There's also a regression-from-3.4.x fix missing. This is all detailed at ​http://gcc.gnu.org/ml/gcc/2005-09/msg00913.html. I'm happy to use 4.0.2 if upstream decide not to do a quick-fix release, though if this is deemed an immensely stupid idea by others then I *may* rethink.