Building 9.2-RELEASE on 10-stable
Andreas Nilsson
andrnils at gmail.com
Sun Nov 10 17:06:11 UTC 2013
On Fri, Nov 8, 2013 at 5:57 PM, Glen Barber <gjb at freebsd.org> wrote:
> On Fri, Nov 08, 2013 at 11:02:12AM +0100, Andreas Nilsson wrote:
> > Now I managed to reproduce the error:
> >
> > A failure has been detected in another branch of the parallel make
> >
> > make[1]: stopped in /src/9.2
> >
> > --- upgrade_checks ---
> >
> > *** [upgrade_checks] Error code 2
> >
> > make: stopped in /src/9.2
> >
> > 1 error
> > make: stopped in /src/9.2
> >
>
> This is one of the races I was talking about.
>
> In some of my tests, I found two things:
>
> 1.) Running 'make make buildworld -j8[...]'* in general would work,
> because make(1) would be built first. That seems to be part of the
> problem here, the toolchain has not yet decided what make(1) to build
> (bmake for fmake), which causes the upgrade_checks phase to fail.
> * Note the 'make make' is not a typo.
>
> 2.) Running the build without '-j' set at all, in most cases, would
> succeed. (Note: even '-j1' would fall over occasionally.)
>
> Glen
>
>
This was indeed tricky...
I think that the check_upgrades target is somewhat broken. On my 10.0-BETA3
box I can't run check_upgrades for either 10.0 or 9.2 sources.
For 10.0:
$ sudo make obj
Password:
/usr/obj/src/10/tools/build/make_check created for
/src/10/tools/build/make_check
andrnils at jailer:/src/10/tools/build/make_check 17:59:16 0
$ sudo make
1..16
ok 1 - C_check # Test of -C flag existence detected no regression.
ok 2 - test_variables # Test variables detected no regression, output
matches.
ok 3 - test_targets # Test targets detected no regression.
ok 4 - sysvmatch # Test sysvmatch detected no regression.
Variable BAR is recursive.
make[1]: stopped in /src/10/tools/build/make_check
ok 5 lhs_expn # Test lhs_expn detected no regression.
ok 6 - notdef # Test notdef detected no regression.
ok 7 - modifiers # Test modifiers detected no regression.
ok 8 arith_expr # Test arith_expr detected no regression.
ok 9 PATH_exists # Test PATH_exists detected no regression.
ok 10 double_quotes # Test double_quotes detected no regression.
ok 11 double_quotes2 # Test double_quotes2 detected no regression.
ok 12 pass_cmd_vars # Test pass_cmd_vars detected no regression.
ok 13 plus_flag # Test plus_flag detected no regression.
ok 14 shell # Test shell detected no regression.
*** Signal 11
Stop.
make[1]: stopped in /src/10/tools/build/make_check
not ok # Test failed: regression detected. See above.
*** Error code 1
Stop.
make[1]: stopped in /src/10/tools/build/make_check
*** Error code 1
Stop.
make: stopped in /src/10/tools/build/make_check
And for 9.2:
$ sudo make obj
/usr/obj/src/9.2/tools/build/make_check created for
/src/9.2/tools/build/make_check
andrnils at jailer:/src/9.2/tools/build/make_check 18:00:11 0
$ sudo make
1..17
ok 1 - C_check # Test of -C flag existence detected no regression.
ok 2 - test_variables # Test variables detected no regression, output
matches.
ok 3 - test_targets # Test targets detected no regression.
ok 4 - sysvmatch # Test sysvmatch detected no regression.
Variable BAR is recursive.
make[1]: stopped in /src/9.2/tools/build/make_check
ok 5 lhs_expn # Test lhs_expn detected no regression.
ok 6 - notdef # Test notdef detected no regression.
ok 7 - modifiers # Test modifiers detected no regression.
make[1]: don't know how to make colons::target. Stop
make[1]: stopped in /src/9.2/tools/build/make_check
not ok # Test failed: regression detected. See above.
*** Error code 1
Stop.
make[1]: stopped in /src/9.2/tools/build/make_check
*** Error code 1
Stop.
make: stopped in /src/9.2/tools/build/make_check
If I disable upgrade_checks dep I get a bit further, but still fails for
libstdc++ on 9.2, running
make make buildworld -s.
By the way, did you mean WITH_GNUXX=1 or WITH_GNUCXX=1 ?
Best regards
Andreas
More information about the freebsd-stable
mailing list