Bugzilla Bug 68395

I recently tried to install a linux32 environment with one of the x86 stages,
but when compiling gcc got the strange message "xgcc: is a directory"gcc:
/usr/lib/gcc/x86_64-pc-linux-gnu/3.4.2/: Is a directory". I traced it down to
GCC_SPECS being set in my amd64 environment, but not unset by linux32 and not
reset by the stage env.d gcc script (which was gcc 3.3.4).
I used stage2-x86-2004.3.20040920.tar.bz2 from the experimental folder, but I
assume the problem is the same when building a 3.4.2 gcc from other stages as
well.

Reproducible: Always
Steps to Reproduce:
1.
2.
3.




Can be solved by "unset GCC_SPECS && emerge gcc"

*** Bug 68394 has been marked as a duplicate of this bug. ***

The bug is more general than the summary suggests as it applies to any switch
from gcc 3.4.2 (perhaps any gcc 3.4.x) to gcc 3.3.x.

I don't think env-update supports commands to unset environment variables, but
could the gcc 3.3.x ebuilds be updated to set GCC_SPECS to something reasonable
instead of not setting it at all?  Otherwise, a switch from gcc 3.4.2 to 3.3.x
using gcc-config, followed by "source /etc/profile" (as the gcc-config script
suggests) leaves the GCC_SPECS environment variable set to its old value, which
can cause some ./configure scripts to fail when emerging certain packages.

GCC_SPECS is no longer set in the default config, and should hopefully leave
the environment completely at some point for the non-default configs. if you've
merged gcc 3.4.2-r2 or 3.4.2-r3 recently, this problem should mostly go away.


*** This bug has been marked as a duplicate of 68799 ***

When updating from sys-devel/gcc-4.3.2-r3 to sys-devel/gcc-4.3.3 everything is
fine. But when switching back from 4.3.3 to 4.3.2-r3, I get this warning:

* gcc-config: Active gcc profile is invalid!                                    
* Switching native-compiler to x86_64-pc-linux-gnu-4.3.2...
* gcc-config: Active gcc profile is invalid!

* Your gcc has a bug with GCC_SPECS.
* Please re-emerge gcc.
* http://bugs.gentoo.org/68395

This goes away when emerging gcc-4.3.2-r3 then a second time.

Should this be this way? Is this a bug?

@M. Anderson: I just found this bug, too, after having some trouble with aspell
and trying to use an older version of gcc. Currently trying to remerge gcc
4.3.3 directly, and see if that makes a difference...

I just got the same message:

* Your gcc has a bug with GCC_SPECS.
* Please re-emerge gcc

when running gcc-config to set up for 4.3.2, after upgrading from
=sys-devel/gcc-4.1.2 to =sys-devel/gcc-4.3.2-r3...on x86.

Things seem to be fine in Portage- and gcc-land, however.

I got the same message as Doug Whitesell in emerge.log for upgrading gcc from
sys-devel/gcc-4.3.3 to sys-devel/gcc-4.3.4 on x86_64.

 * Your gcc has a bug with GCC_SPECS.
 * Please re-emerge gcc

 Things seem to be fine in Portage- and gcc-land here, too.
gcc-config -l
 [1] mingw32-4.4.0 *

 [2] x86_64-pc-linux-gnu-4.3.4 *

Do we really need to reemerge gcc?

I've also run into this when upgrading sys-devel/gcc-4.3.2 to
sys-devel/gcc-4.3.4 on x86_64

>>> Installing (1 of 1) sys-devel/gcc-4.3.4
 * Running 'fix_libtool_files.sh 4.3.2'
* Scanning libtool files for hardcoded gcc library paths...
* gcc-config: Active gcc profile is invalid!
gcc-config: error: could not run/locate 'gcc'
:0: assertion failed: (gcc -dumpversion) | getline NEWVER)
* gcc-config: Active gcc profile is invalid!
* Switching native-compiler to x86_64-pc-linux-gnu-4.3.4...
* gcc-config: Active gcc profile is invalid!

* Your gcc has a bug with GCC_SPECS.
* Please re-emerge gcc.
* http://bugs.gentoo.org/68395


But gcc was upgraded correctly anyway

# gcc --version
gcc (Gentoo 4.3.4 p1.0, pie-10.1.5) 4.3.4

AFAICS (regarding comments #4 to #8) gcc-config says this when switching from
an invalid profile (e.g. because you just uninstalled the active gcc version)
to another one. When you run gcc-config again, the new profile is active and
working, and no warning is displayed.

I consider this a bit of a confusing uglyness in gcc-config, maybe we should
file another bug?

If that's really the case, filing another bug sounds great (I just ran into
this, too, on a couple of boxes).

(In reply to comment #10)
> If that's really the case, filing another bug sounds great (I just ran into
> this, too, on a couple of boxes).

Filed a new bug on this: bug #290437.

People who came here because your gcc-config told you to visit this page:
Probably you don't need to re-emerge gcc, have a look at bug #290437 and see
whether it applies to you.