#2856 dejagnu in koji started to "spawn failed" everything
Closed: Fixed None Opened 12 years ago by jankratochvil.

= phenomenon =

gcc build always runs automatically a testsuite from %check. Now the testsuite fails to run.

= reason =

The last build was https://koji.fedoraproject.org/koji/buildinfo?buildID=246208 on 2011-06-03 which was successful.
Now my scratch build https://koji.fedoraproject.org/koji/taskinfo?taskID=3161441 with (almost) the same .src.rpm on 2011-06-25 fails.

{{{
-Test Run By mockbuild on Fri Jun 3 16:00:49 2011
+Test Run By mockbuild on Sat Jun 25 23:36:35 2011

spawn -ignore SIGHUP /builddir/build/BUILD/gcc-4.6.0-20110603/obj-x86_64-redhat-linux/gcc/xgcc -B/builddir/build/BUILD/gcc-4.6.0-20110603/obj-x86_64-redhat-linux/gcc/ -O0 -w -c -o 20000105-1.o /builddir/build/BUILD/gcc-4.6.0-20110603/gcc/testsuite/gcc.c-torture/compile/20000105-1.c^M
-PASS: gcc.c-torture/compile/20000105-1.c -O0 (test for excess errors)
+compiler exited with status -1
+output is:
+spawn failed
+FAIL: gcc.c-torture/compile/20000105-1.c -O0 (test for excess errors)
}}}

= recommendation =

Maybe mock Bug 701891 mentioning new mock 1.1.11 may be related.


Re-verified on original .src.rpm koji rebuild: {{{fedpkg build --scratch}}}[[BR]]
https://koji.fedoraproject.org/koji/taskinfo?taskID=3161563

I'm not seeing the "compiler exited with status -1" output in your scratch build.log. Are you sure you were looking at the same output ?

I do see:
Couldn't determine version of /builddir/build/BUILD/gcc-4.6.0-20110603/obj-x86_64-redhat-linux/gcc/xgcc: spawn failed

But I don't see much that points to koji here. Does it fail the same way in a local mock build? It sounds like a code/package/patch issue to me...

Replying to [comment:2 kevin]:

I'm not seeing the "compiler exited with status -1" output in your scratch build.log.

You need to uudecode it first and tar xjf it and see for example gcc.log there.

But I don't see much that points to koji here.

It works with local rpmbuild. Please compare the regression on these two tasks, both come from the same fedpkg verrel GIT hash:[[BR]]
https://koji.fedoraproject.org/koji/taskinfo?taskID=3108168 [[BR]]
https://koji.fedoraproject.org/koji/taskinfo?taskID=3161563

Does it fail the same way in a local mock build? It sounds like a code/package/patch issue
to me...

The exactly same Koji source rebuilt in Koji had correct testsuite before a month and now the testsuite produces "spawn failed" only.

So, can you try building locally with mock 1.1.9 mock 1.1.10 and mock 1.1.11 and see if this is a mock issue? (Builders have 1.1.10 currently).

I can try here, but might be a bit before I get time...

bukaj: jkratoch: I've just tested mock on f15 and it doesn't work either[[BR]]
bukaj: jkratoch: so it is a recent mock bug, related to /dev/pts mount again[[BR]]
bukaj: jkratoch: doing mount -t devpts none /dev/pts in the mock chroot makes it work again[[BR]]
bukaj: mock-1.1.10-1.fc15.noarch

On updated F14 x86_64 both current mock-1.1.10-1.fc14.noarch and downgraded mock-1.1.9-1.fc14.noarch have the same "spawn failed" problem on:[[BR]]
{{{mock -r fedora-15-x86_64 --rebuild gcc-4.6.0-10.fc15.src.rpm}}}

mock-1.1.11-1.fc15 ought to fix this (it removed ,newinstance ; or, hopefully 1.1.12 which readds it but makes /dev/ptmx a symlink to pts/ptmx if ,newinstance is used).

See #683111 for more details.

mock-1.1.11-1.el6 has been pushed to stable, could the build servers be updated please?

mock has been downgraded on all the builders. Can you see if this solves the issue for you?

Please do reopen if it doesn't.

Login to comment on this ticket.

Metadata