Package: src:capnproto Version: 0.7.0-3 Severity: normal Tags: upstream Dear Maintainer,
When trying to build a stretch backports of capnproto, I hit a build test failure: [ TEST ] kj/filesystem-disk-test.c++:835: DiskFile holes kj/filesystem-disk-test.c++:889: failed: expected file->stat().spaceUsed == meta.spaceUsed stack: 55e8f965ca5c 55e8f96e4473 7f1b104b727f 7f1b102612d8 7f1b104b863d 7f1b104b9670 7f1b102929df 7f1b102946a6 7f1b102612d8 7f1b10294864 7f1b104b601e 7f1b0f6182e0 55e8f96444d9 [ FAIL ] kj/filesystem-disk-test.c++:835: DiskFile holes (248617 μs) I don't know what causes this, my only idea is BTRFS, as in the title, though that's a pure guess. I build with sbuild on a BTRFS snapshot of the clean build chroot. As a workaround, I added the following quilt patch: --- capnproto-0.7.0.orig/src/kj/filesystem-disk-test.c++ +++ capnproto-0.7.0/src/kj/filesystem-disk-test.c++ @@ -886,7 +886,7 @@ file->truncate(1 << 21); file->datasync(); - KJ_EXPECT(file->stat().spaceUsed == meta.spaceUsed); + // Fails on BTRFS: KJ_EXPECT(file->stat().spaceUsed == meta.spaceUsed); KJ_EXPECT(file->read(1 << 20, buf) == 7); KJ_EXPECT(StringPtr(reinterpret_cast<char*>(buf), 6) == "foobar"); which resulted in a successful build. I'm not sure I'll have a chance to dig into this, but wanted to make sure the information isn't lost. I'm not tagging with patch, because it's based on a guess only. -- Regards, Feri.